The technology is real, and the demand for computing power is also very real. Businesses are spending heavily on AI, consumers are using it at unprecedented scale, and the capabilities of the models continue to improve at a dizzying pace. All good on that front.

The problem though, is the financial structure being built around that demand. It’s showing up in two main ways.

The first is capital. The same handful of companies increasingly sit on both sides of the industry’s biggest transactions: investor, supplier and customer at the same time. Nvidia has committed up to $100 billion to OpenAI while supplying the chips that power its models, and holds a stake in CoreWeave, which sells that same computing power back to both OpenAI and Microsoft — while Microsoft is one of OpenAI’s largest shareholders and one of Nvidia’s biggest customers.

The list goes on and on, but I think you get the gist.

The second is the product itself. ChatGPT, Claude, Gemini and their competitors are becoming dramatically cheaper to use with every new generation, while open-source models are closing the gap with the frontier labs. Capability is rising, but the price of intelligence is falling.

That combination creates an awkward tension (for the frontier labs at least). The industry is spending extraordinary amounts of money building infrastructure for a product whose marginal price is moving steadily towards zero. Will the likes of Anthropic and OpenAI eventually struggle to sell their models at a high enough price to justify current expenditure levels?

Don’t get me wrong, the demand for computing power is undeniable, the technology is improving rapidly, and many of the applications being built on top of it will be enormously valuable. There is no question about that. But the financing structure underneath the boom is becoming increasingly worrisome.

The bubble, in our view, is not AI itself. It is the assumption that the current economics of AI can continue indefinitely.

Here’s the case for why we think that assumption breaks and why it doesn’t change how we invest at Momentous.

The self-feeding loop

The circular spending which is increasingly being covered publicly is built on a very narrow customer base.

That structure can work totally fine while growth remains strong. The problem is what happens if and when one of the central customers slows down.

OpenAI and Anthropic are expected to account for an increasingly large share of cloud revenue, despite neither company being consistently profitable. UBS estimates the pair will make up 27% of Google Cloud’s revenue this year, rising above 48% next year. Microsoft’s cloud division, meanwhile, has already become heavily exposed to OpenAI. In fact, 69% of its growth in 2025 came from that one customer alone, and growth would have been a meager 8% without it. And Google has built a very similar relationship with Anthropic. In effect, some of the world’s largest technology companies are relying on a small number of AI labs to justify an enormous amount of infrastructure spending.

Let’s play that forward. If OpenAI ever pulled back (i.e. scaled down its compute spending, lost ground to a rival, or simply grew slower than forecast) Google Cloud and Microsoft Azure would see a meaningful chunk of their revenues evaporate. Public markets, which price the stocks on the assumption that growth continues at pace, would see a considerable correction and broader market sell-off on fears of a bubble.

Under scrutiny, the numbers quickly become difficult to reconcile. One widely cited estimate put the data-centre capacity now being built at around 190 gigawatts, requiring something like $1.6 trillion in annual revenue to justify the investment. These figures aren’t audited, but they give you an idea of the scale we’re talking about (and betting on). OpenAI itself reportedly posted $13.1 billion of revenue in 2025 against a $38.5 billion net loss, while committing hundreds of billions of dollars to future computing capacity.

And much of the financing isn’t immediately visible. A lot of it runs through leases and separate entities that build and own data centres on a hyperscaler’s behalf — arrangements that work more like renting than borrowing, so the obligation is just as real but doesn’t always show up as debt on the books. Add bonds, leases and this kind of hidden borrowing together, and the industry’s total obligations come out to somewhere near $1 trillion, per reporting from Nikkei Asia and Moody’s — a meaningful chunk of it ultimately funded by pension and insurance money, in what the Bank for International Settlements calls “shadow borrowing.”

This doesn’t prove that the spending is wasteful by any means. There are very legitimate reasons to believe demand for compute will continue to grow. There’s also a structural advantage specific to the hyperscalers that doesn’t apply to OpenAI or Anthropic: Google and Microsoft aren’t betting the house on AI working as a standalone product. Rather, they’re layering it onto businesses that are already enormous and already profitable. In 2025, Google and Microsoft posted net profits of $132B and $102B respectively. Google’s roughly 90% share of global search isn’t going anywhere either; Gemini and AI Overviews are being used to cleverly defend that position. If the AI trade corrects, both companies still own the businesses AI was built on top of.

Simply put, the point is that the financing structure leaves very little room for error. When that error eventually comes (in the shape of something subtle like major AI labs missing expectations, or cutting capex), investors will question the assumptions supporting the companies around it. And spoiler alert… the losers won’t be the hyperscalers (Amazon, Google, etc.).

The model layer has no moat

This becomes more interesting when you look at what is happening to the economics of the models themselves.

The best AI models used to come in one flavour: closed. You paid a company like OpenAI or Anthropic for access, and the model itself stayed private (centralised and controlled by the company that owns it). Increasingly, that’s no longer the only option. “Open-weight” models — versions anyone can download, run on their own servers, and modify freely, from labs like China’s DeepSeek and Moonshot, or tech companies like Meta and Nvidia — have closed the performance gap dramatically, especially on the tasks businesses actually pay for, such as coding. For many buyers, whether a model is “open” or “closed” now matters less than whether it’s good enough and cheap enough. Add the appeal of running your own data privately and securely on your own servers (rather than sending it to someone else’s through the cloud), and open-source starts looking like the more attractive option. Particularly for businesses.

When models become interchangeable on quality and performance, price starts to matter more, and the price gap is enormous as it stands. AI companies charge based on how much text a model reads and writes, measured in “tokens” (roughly chunks of a word). DeepSeek’s models cost as little as $0.28 to $0.87 for every million words of output; the best Western models charge $15 to $25 for the same amount — 20 to 30 times more, for performance that Epoch AI, an independent AI research group, puts only about four months behind. For the vast majority of use cases, that remaining gap doesn’t really matter though — it’s only relevant at the frontier, for things like drug discovery or advanced cybersecurity research that 99% of people and businesses will never ever touch. For everything else, the models are already interchangeable, and price will decide the winner.

Across the board though, the technology keeps improving fast, and every generation makes it cheaper to actually run these models day to day. More competitors enter the market. Open-source models get easier to deploy. Hardware improves. And the result is that the cost of intelligence keeps falling.

That is fantastic for businesses building on top of AI. It is much less attractive for the companies whose valuations depend on maintaining pricing power at the model layer (i.e. Anthropic, OpenAI).

The AI trade is often compared against the dot-com boom or even cloud computing. But there is a more useful historical analogy here: Mobile data. A sharp point Benedict Evans made well in this piece.

TL;DR: As cellular networks improved, the amount of data people consumed exploded. The underlying network became enormously important, but much of the economic value accrued to the businesses built on top of it (Instagram, Uber, Netflix, …) rather than to the network itself. Telcos built the infrastructure and captured comparatively little of the upside; their stocks have been notoriously flat for two decades. The network simply became infrastructure.

AI could follow a similar trajectory where intelligence becomes abundant and cheap, while the value accrues to the software layer (specialised workflows, proprietary data sets, vertical business solutions). A handful of companies are currently aggressively financing the infrastructure before we know where the value will settle.

Public and private markets running one bet

The Magnificent Seven — the seven biggest US tech companies, including Nvidia, Microsoft, Google and Amazon — now make up roughly a third of the value of the entire S&P 500, the index tracking America’s 500 largest public companies. Nvidia alone accounts for around 7.5% of it. AI-related investment accounted for the large majority of US GDP growth in the first quarter of 2026, while spending on data centres, hardware and networking reached around 1.4% of GDP. The productivity gains that are supposed to justify this haven’t shown up at anything close to that scale, and companies are massively exceeding their AI budgets rather than saving on them.

Public markets are also suddenly being asked to absorb an unusual amount of new AI-linked equity at once. SpaceX’s IPO — the largest in history, at a $1.77 trillion valuation — will likely be followed by even larger Anthropic and OpenAI listings, all while the Mag 7 simultaneously ask investors to finance their own buildout. Everyone is drawing from the same pool of capital, which is okay while expectations stay high but could lead to a steep correction once that stops.

South Korea just showed us a miniature version of what that might look like. Samsung and SK Hynix, the country’s two big AI chip suppliers, came to make up more than 60% of the Kospi (South Korea’s main stock market index). Leveraged retail money — ordinary investors borrowing to buy more than they could otherwise afford — continued to pile in, and when US chip stocks wobbled in late June (purely on investor sentiment, nothing more), the index shed 16% in two days…

Private markets carry the same concentration. AI accounted for 61% of global venture funding in 2025, rising to 79% by Q1 2026, with three deals — OpenAI, Anthropic and xAI — taking two-thirds of it. At the seed stage, AI companies commanded a 42% valuation premium over non-AI peers, even as deal counts fell and capital concentrated with a smaller group of brand-name managers.

When the top of the market reprices, it resets the comparable valuation every investor uses underwriting a seed or Series A round — whether or not that company has anything to do with AI.

That’s why we think a correction is due. Even though the demand is very real, the financing structure around it has become too dependent on a small number of companies continuing to grow at extraordinary rates, and we don’t need to know the exact trigger to see how little room for error that leaves.

And that brings us to the part of the market where we think the implications are most interesting.

What doesn’t compress

The Human Potential Stack sits outside this particular loop. Momentous’ performance, wellness and longevity sectors are not priced on the assumption that AI infrastructure spending needs to compound indefinitely, or that frontier models need to command a price premium. In fact, they benefit from AI becoming cheaper.

As digital intelligence becomes abundant, anything that cannot be digitised becomes relatively more scarce. Sport is the clearest example. The LA Lakers recently sold to Bob Iger and Josh Kushner at a record $12.5 billion valuation — barely a year after Mark Walter bought control at $10 billion. Meanwhile, Liverpool has just sold about 30% of the club to the 1892 Holdings consortium at a record $7B+ valuation, with Jeff Bezos as lead investor. Investors haven’t suddenly woken up and decided to pay that price because these sports are becoming more technologically sophisticated. They’re paying for something technology cannot reproduce.

You cannot download a live sporting event, or automate the experience of being in a stadium with thousands of other people. Technology can make digital content abundant, but it cannot make scarce physical experiences abundant. And we believe AI will only increase the value of that scarcity.

The same is true for wellness and longevity, arguably with an even more durable underlying demand driver. The global wellness economy reached $6.8 trillion in 2024 and is forecast to approach $9.8 trillion by 2029. These markets are driven by demographics and behaviour (growth that is completely separate from a hyperscaler’s next earnings call). People want to perform better. They want to feel better. They want to live longer. And they want to do it with other people, in the real world.

That said, AI will still play an enormous role here as the underlying technology, improving diagnostics, personalising interventions, automating monitoring and accelerating discovery. Businesses operating in health and sports will use cheaper intelligence as an input to drive efficiency gains, while selling something much harder to commoditise.

AI is incredibly useful. Clearly. The question is whether a small group of companies selling compute to each other can continue growing fast enough to justify the enormous amount of infrastructure being built around them. We don’t think they can forever, and for Momentous, that is not a problem we need to solve.

Our portfolio is built around the stadium seat, the healthier body, the stronger performance, the extra year of life. In other words, the things that people actually get to experience and enjoy.

As artificial intelligence becomes cheaper, being human becomes more valuable.

Ours are not necessarily more talented. But they are more intentional about staying good.

Novak Djokovic did not just extend his career; he extended his prime. Cristiano Ronaldo has turned self-maintenance into a professional discipline in its own right. LeBron James is still performing at an age where most players have either retired or reframed themselves as “experienced veterans.” Serena Williams and Allyson Felix have shown the same infrastructure can survive being reset entirely — by childbirth, by injury, by four years away from competition — and still come back for more.

This is what deliberate consistency sustained over decades looks like.

That is why longevity has become such an interesting category in performance. Not because everyone (aka Bryan Johnson) suddenly wants to become immortal. But because longevity is what happens when you start treating the body like an adaptive system that requires maintenance, feedback, and intelligent constraint.

Many people view longevity as extending life at all costs. It’s not. I believe longevity to be about extending capability — staying sharp, strong, and effective for longer than the average trajectory would suggest is possible. Therefore, lifespan (how long we live) always has to go hand in hand with healthspan (the number of years we live in good health).

Genius Without a System

Before diving into the greats of this generation, let’s not forget that none of this longevity talk mattered to the athletes of a previous era, and two of the greatest prove exactly how little it mattered. Well… until it did.

Diego Maradona won the 1986 World Cup almost single-HANDEDly (if you know you know), carrying a modest Argentina squad through a seven-match tournament of individual brilliance that owed nothing to sports science, sleep architecture, or a full-time nutritionist. Johan Cruyff redefined what football could look like tactically while also smoking through half-time team talks, a habit so embedded in his identity that opposing benches could smell him coming. Neither man built infrastructure around his talent. Neither seemed to think he needed to.

However, the bill came due later, and it was steep. Maradona, who by the looks of it would have enjoyed today’s Enhanced Games, failed a drug test for cocaine in 1991 and was banned from playing for fifteen months. This marked the start of a decline that shortened his career and, eventually, his life; he died in 2020 at 60, following decades of addiction. Cruyff smoked through his playing career and into his time as Barcelona’s coach, then needed emergency double bypass surgery in 1991 to survive it. He quit that day and fronted an anti-smoking campaign afterwards. Unfortunately, he still died of lung cancer in 2016, aged 68. Both men reshaped their sport permanently, but neither got the second act that this generation’s greats are still writing.

This is not an argument that talent needs discipline to matter, or that Maradona and Cruyff would have been greater with a better spreadsheet. In fact, their fast life and lack of spreadsheets probably gave them unlimited aura.

It is more so a comment on what each model can promise. Genius without infrastructure can still produce the best years anyone has ever seen. What it cannot promise is more of them. The current generation is not necessarily more gifted than the last one, it has simply worked out how to keep the gift alive for longer.

Discipline by Design

In discussions, longevity gets framed as a protocol — a stack of supplements, a dashboard of biomarkers, a perfectly optimised diet — as if the right combination of inputs guarantees the outcome. But athletes who actually last do not win because of one magic variable. They win because they construct environments where the right behaviours happen by default, and the destructive ones are engineered out. One relies on fluctuating motivation. The other relies on compounding infrastructure.

Djokovic and Ronaldo show what that infrastructure looks like in practice, built from opposite directions. With Djokovic, the obvious story is titles — twenty-four Grand Slams, the most in men’s tennis history, and even with Alcaraz and Sinner playing like gods, you wouldn’t put it past him to win number 25 at the age of 39. Djokovic reframed inflammation, stress, sleep, and nutrition as competitive variables rather than wellness footnotes, then built his career around subtraction through a plant-based diet, gluten elimination, and breathwork.

Ronaldo arrived at the same principle from the opposite direction: automation rather than subtraction. Recovery is scheduled with the rigour of training. Nutrition is close to mechanical. Sleep is structured around multiple cycles rather than one nightly block. Every decision that depends on motivation is one that might not survive contact with pressure, so he has engineered motivation out of the equation entirely.

LeBron adds a third dimension to the same principle: decline is rarely one dramatic break. It is accumulation — micro-injuries that never fully heal, travel fatigue compounding across an 82-game season, chronic sleep debt eroding reaction time. None of it ends a career alone, but the accumulation is why most players are done by 35. LeBron has built an obsessive buffer against that accumulation instead, reportedly to the tune of a million dollars a year on cryotherapy, hyperbaric chambers, a full-time nutritionist, and meticulous sleep tracking. He has stayed relevant “simply” by being harder to break than everyone else.

Djokovic, Ronaldo, and LeBron each used different mechanisms but very much with the same outcome in mind, and none of them are relying solely on willpower to show up.

Comebacks at the Highest Level

Serena Williams and Allyson Felix prove the same infrastructure has to survive the body resetting itself entirely, as they refused to choose between their careers and having a family. They’ve shown that you can have both.

Williams won the 2017 Australian Open around eight weeks pregnant, then delivered Olympia by emergency Caesarean that September (an ordeal complicated by a pulmonary embolism that left her bedridden for six weeks).

She already returned to the tour within a year, reaching four grand slam finals and winning the 2020 Auckland Open before eventually retiring in 2022. The durability behind her recent 2026 Wimbledon return was as much metabolic as it was physical. After her second pregnancy left her blood sugar and joints in worse shape than at any point in her career, she started wearing a continuous glucose monitor to see in real time how food, sleep, and stress moved her numbers, and began taking a GLP-1 medication that brought her blood sugar back to a normal range and took off more than thirty pounds. Four years after what looked like a final retirement, she stepped back onto Centre Court at 44 to play Wimbledon singles for the first time since 2022, taking a set off a player less than half her age before finally losing in three. Age did ultimately matter… but what a legend!

Felix’s athletic reset came from an emergency Caesarean at 32 weeks, brought on by severe pre-eclampsia that threatened her life alongside her daughter’s. She returned to win Olympic gold in Tokyo and became the most decorated athlete in US track and field history, in large part by treating recovery as seriously as training, through consistent sleep, daily soft-tissue work, and a rule of progressing only once pain-free rather than on a fixed calendar. That discipline once meant choosing months of pool-based rehab over surgery for partially torn ankle ligaments rather than risk a shortcut back. Now 40, she has announced a second comeback titled “Project Six”: a bid to make the 2028 Los Angeles Games her sixth Olympics, which would make her the first American track and field athlete ever to reach that number.

Neither treated motherhood, or age, as a career interruption to recover from. They’ve shown tremendous durability, grit, and patience — rebuilding on the body’s timeline instead of the ego’s, and coming back only once ready rather than as soon as possible.

The Founder Translation

Every part of this translates to startups. Founders are not elite athletes, but their constraints are still biological, and the companies that last are not the ones that move fastest through a single hype cycle or bubble. They are the ones built to survive the next one, and the one after that.

Startups accumulate their own version of chronic inflammation: the shortcut that seemed harmless at the time, the process nobody owns because everyone assumed someone else would, the technical debt quietly compounding in the background. None of it kills a company on its own; it just makes every subsequent decision slightly worse.

You cannot outwork biology indefinitely. It collects what it is owed whether or not that is convenient — showing up as reduced clarity in a critical decision, or a cultural tone set by exhaustion that quietly becomes how the team works.

Longevity, for a founder, means staying sharp long enough for compounding to matter. Strategy, relationships, and reputation all compound, but none of it does if the operator, or the company, burns out first.

A final thought. None of this means optimising everything, all the time, at all costs. Drinks at the pub with friends, a greasy burger, or a fat milkshake are not failures of the system. They are very much part of it.

Although it can often come across that way, joy is not the enemy of longevity. Chronic joylessness probably is though. The athletes above did not build systems in order to live less; they built them in order to live and perform more fully, for longer. The goal is a full life worth sustaining, not just a fit body you’ve miserably managed into old age.

That camera angle was a useful way to see what was actually happening at the tournament. The opener alone needed every bit of officiating support available with three red cards before the final whistle, the most in any World Cup opening match. The tech being used to support the match officials didn’t turn into a conversation about AI replacing referees or predicting outcomes (though “prediction” [read betting] markets did have a successful tournament as well).

The story was narrower and much more interesting than that: a layered stack of computer vision, sensors, and generative models, built to make football’s most contested decisions visible, explainable, and fast enough to matter while the match was still live.

Moving at the speed of the game

Semi-automated offside technology debuted in Qatar in 2022, but it only flagged a player once they were more than 50cm offside, and even then the alert routed through the VAR room before anyone on the pitch heard about it. That delay was why assistant referees were trained to hold their flag and let play develop. An unnecessary step that slowed down the game and dampened the excitement. It also contributed to Nottingham Forest striker Taiwo Awoniyi’s serious abdominal injury in 2025, one of the incidents officials pointed to when arguing for change. This year, FIFA thankfully tightened the threshold to 10cm and rewired the alert to go straight into the official’s earpiece. The flag went up the moment a violation was clear, not after a video booth confirmed it.

The system still couldn’t do everything though. It only caught positional offside, not interference, and it sat out the closest calls entirely. Underneath it sat another piece of hardware doing a lot of the actual work. The Adidas Trionda (official match ball) carried a motion sensor recording 500 times a second, precise enough to mark the exact instant of a kick or a handball touch and feed that timestamp straight into the system. Paired with it was something equally impressive. Every player got a precise 3D body scan before the tournament, accurate enough to track them through a crowded box or a collision.

Any football fan will tell you that offside has always been less a measurement problem than a trust problem. Fans don’t reject calls because they’re wrong, they reject them because they’re inscrutable. A clean 3D reconstruction on the broadcast went a long way in providing that trust factor. FIFA pushed that same instinct into the stadium itself, ensuring that after a VAR review, the referee now explained the call out loud over the public address system, instead of leaving thousands of people in the stands to guess from the players’ reactions. This has been a common practice across various other sports for years, but it’s good to see that football has now caught up.

The importance here is that FIFA didn’t just make the system more accurate, it made the accuracy legible to the people it affected. A verdict without a trustworthy and verifiable explanation (however accurate it may have been) would not have had the same impact on the game. A faster, more correct decision nobody understands still loses the room. But a decision people can see for themselves doesn’t need defending at all. Well, maybe a little, it’s still passionate football fans we’re dealing with here!

That lesson held in reverse, too. During Qatar’s group match against Switzerland, the system that generates the offside graphic crashed before a contested penalty decision, leaving FIFA with no animation to show and only a statement insisting the underlying call was correct. The explanation arrived without the evidence, and that wasn’t enough. Pundits accused FIFA of hiding something even though the call may well have been clean. The failure wasn’t in the offside decision itself, but in the layer built to make the decision visible. The layer people actually trust.

Analysis for all

Lenovo’s partnership saw them build FIFA AI Pro, a generative assistant trained on FIFA’s own match data, which was given to all 48 competing teams equally. It answered questions about opponents and tactics in text, video, and 3D visualisation, in whatever language a federation needed.

The gap between a team with a full analytics department and one without has long been one of the clearest predictors of how far a squad could go. Handing every federation the same digital assistant closed that preparation gap, giving teams that couldn’t afford an analytics department access to one.

Fans inside stadiums equally had access to their own analysis. They could point their phones at the pitch through the FIFA+ app to see an augmented reality overlay showing real-time player names, speeds, and physical intensity. The same data layer powering officiating decisions, surfaced directly into the stands.

The unglamorous bottleneck

A 48-team format meant 104 matches across three countries. A level of volume that could have easily broken the old broadcast playbook. Verizon deployed private 5G networks at all 16 venues, with capacity at the 11 US stadiums upgraded by up to five times to handle the volume of fans streaming and uploading simultaneously. Lenovo ran near-real-time infrastructure out of Dallas to log, tag, and clip footage fast enough to keep up, and the Referee View stabilisation alone cut visual jitter by roughly half.

None of those individually made major headlines, but combined they did make the world cup watchable at global scale. It was the part of the stack that got the least credit while enabling everything else on top of it.

The off-pitch trust problem

The same month the tournament opened, a clip of a Boston Dynamics robot dog patrolling a Dallas concourse went viral, captioned as proof that fans were being secretly face-scanned. Boston Dynamics and FIFA both confirmed the robots had no facial recognition capability; their job was perimeter checks and flagging suspicious packages to human security. Still eerily close to a Black Mirror episode…

The irony was that the real version of the facial recognition fear did exist, just not where anyone was looking. Several US stadiums ran actual AI-powered facial recognition for entry and payment throughout the tournament. The public aimed its suspicion at the unsettling machine that couldn’t see faces and missed the mundane turnstile that discreetly could.

So what?

Strip away the broadcast graphics and the World Cup ran the same stack every company building in elite performance eventually has to reckon with: sensing, real-time inference, decision support, explainability, distribution, and governance, compressed into 39 days and 104 matches.

What separated the systems that worked from the ones that were just noise? Every one of them tried to make a decision faster and more legible to the people who had to trust it.

That’s the discipline worth carrying out of this tournament. The products that last aren’t the ones forecasting what will happen. They’re the ones that make a decision actionable and trustworthy.

Matteo Franceschetti arrived at that premise the long way. A former lawyer and competitive athlete, he became fixated on a simple, uncomfortable fact: he spent a third of his life in an environment he had no control over, and that environment was working against him. The founding insight wasn’t complex, but it required looking directly at what the mattress industry had spent decades ignoring: temperature is the primary biological driver of sleep quality, and nobody was actively managing it. The body needs to drop its core temperature to initiate sleep, and to cycle through specific thermal ranges to move between deep sleep, REM, and lighter stages. A standard, static foam rectangle, heated by body mass all night, does the opposite of what the body actually needs. Franceschetti partnered with engineer Massimo Bassi and PR strategist Alexandra Zatarain, and after an initial rejection from Y Combinator, famously returned to build a fully functional bedroom inside the YC interview room and secured admission the second time of asking.

Today, Eight Sleep operates at institutional scale. The Pod has surpassed $500 million in cumulative sales, with revenue growing tenfold since its introduction. In March 2026, a $50 million strategic round led by Tether Investments valued the company at $1.5 billion, up from $1 billion in August 2025, and $500 million at its Series C in 2021. Total funding raised stands at over $310 million. The company reached free cash flow positivity in 2025 and now ships to 34 countries. Eight Sleep is no longer a hardware startup. It is in the process of becoming something considerably more consequential: a predictive health intelligence platform anchored in the bedroom.

Hardware as a Clinical Lab

Eight Sleep confronted hardware risk first. Before there was a product worth selling, there was a genuinely difficult engineering problem: building a system capable of pumping water through a mattress cover continuously, without leaking, without failing under human body weight, and without disturbing the sleep it was designed to improve. There was no question of whether people wanted to sleep better. The open question was whether the company could build a physical system precise enough to validate the underlying science in the first place.

Validating that science required clinical investment. Eight Sleep embedded sensors across the Pod cover to continuously track heart rate, respiratory rate, and heart rate variability (HRV), without requiring the user to wear anything. An early peer-reviewed study of 54 participants crucially demonstrated that the Pod increased deep sleep by 14 minutes and REM sleep by 9 minutes in specific cohorts, lowered resting heart rate by two percent, and increased HRV by seven percent. These were the foundation on which everything else was built.

The intended outcome of that investment was verifiable sleep improvement. The unintended outcome was vastly more valuable. By occupying a permanent position in the bedroom, Eight Sleep created a continuous health data collection system that operates every night, without friction, without compliance decay, and without the user having to remember to wear a device. The company now trains its predictive models on over one billion hours of real-world sleep data. That dataset is not the product of a research study. It is the compounding result of a thoughtful consumer hardware business.

Building physical products is the required path to capture proprietary longitudinal biological data.

Rest as Athletic Performance

Eight Sleep didn’t just need to sell an expensive mattress cover. It needed to make spending thousands of dollars on a mattress cover feel rational (aspirational, even). That required a cultural reframe that reached far beyond any simple marketing campaign.

The company identified its entry point precisely. High-performers, elite athletes, and biohackers were the required early adopters — because of their spending power, but more importantly because of the cultural signal they carried. These were people who already tracked their nutrition, their training load, and their recovery metrics. Sleep, in their context, wasn’t a by product — it was the most underoptimised variable in an already-optimised system. Eight Sleep adopted the language of sleep fitness, positioning the Pod as a performance tool rather than another rectangle of expensive foam.

The credibility was built deliberately. Neuroscientist Dr. Andrew Huberman, sleep researcher Dr. Matthew Walker, and physician Peter Attia joined the Scientific Advisory Board, anchoring the platform’s claims in scientific legitimacy (though the latter has since been removed due to his name’s repeated appearance in the Epstein files). Top tech executives including Elon Musk and Mark Zuckerberg also adopted the product publicly which did a lot for the brand. Today, Eight Sleep has built a high calibre roster of more than 300 professional athletes who use the Pod for recovery, including Lewis Hamilton, NBA forward Jimmy Butler, and Yankees outfielder Aaron Judge.

Meanwhile F1 driver Charles Leclerc and tennis player Taylor Fritz both joined as athlete ambassadors and investors, putting their own capital behind the product they use nightly. Bryan Johnson, whose extreme longevity routines attract outsized media attention, equally made Eight Sleep a visible fixture in his regimen.

Each of these associations served a dual purpose.

Functionally, elite users created a brutal feedback environment — people whose performance depends on genuine physiological edge don’t stay loyal to products that don’t deliver. Their continued advocacy was a form of invaluable product validation.

Culturally, they repositioned the category entirely. Once the Pod sat in the bedroom of a Formula 1 driver and a Silicon Valley billionaire, the question of whether the price was justified shifted from financial to aspirational.

Category creation requires identifying the most demanding users and making their obsession aspirational. Nail that and the awareness naturally follows.

Paying for the Pod Twice: Hardware and Subscription Economics

Eight Sleep generates revenue through a model that many consumer hardware companies have attempted and few have executed cleanly. A high-margin upfront hardware sale paired with a recurring software subscription.

The base product is the Pod Cover, starting at approximately $2,000. The full mattress system (the cover plus Eight Sleep’s own foundation) pushes past $3,000. The Pod 5 Ultra, launched in May 2025 as the flagship configuration with an adjustable base enabling automatic head elevation and a hydro-powered blanket, starts above $5,800.

Beyond the hardware, customers are required to subscribe to Autopilot — the software layer that enables automatic temperature adjustments, biometric tracking, and AI-driven sleep coaching — at $17, $25, or $33 per month depending on the tier. Note that the subscription is not optional. Access to the full functionality of the hardware depends on it.

Adjacent revenue streams further extend the ecosystem. The hydro-powered blanket and a temperature-controlled pillow cover each retail at roughly $1,000. The company is also testing a line of sleep-supporting supplements, a move that signals a broader ambition to own more of the user’s sleep-adjacent spending.

Speccing out the product with all the bells and whistles will take the price well over $7,000.

A bargain.

As you may have guessed, the subscription model is not well received. Users routinely raise the question of paying monthly fees to access hardware they have already purchased outright. The dependency on cloud infrastructure has produced real operational exposure: an AWS outage in October 2025 left Eight Sleep devices stuck wherever they happened to be when the outage hit — some users’ beds locked at 43°C, others stranded in inclined positions, with no way to adjust temperature or flatten the base. Eight Sleep shipped an offline Bluetooth “outage mode” shortly after, but the incident exposed the brutal reality that a physical product costing upwards of $2,000 had no functioning manual fallback.

Mandatory subscriptions on physical goods generate the capital and cash flow required to scale hardware businesses and develop AI capabilities. They also generate consumer friction.

The Real Value Exchange

Nobody buys an Eight Sleep Pod because they want a mattress cover. Certainly not at that price. Understanding what they are actually purchasing requires looking past the feature list.

The first thing couples buy is the elimination of thermal conflict. The Pod’s dual-zone architecture allows each side of the bed to operate at a completely independent temperature — one partner sleeping cool while the other sleeps warm. For the significant portion of couples whose sleep is disrupted by temperature disagreement, this is a solution to a critical problem, and it makes the individual purchase decision a shared one.

The second thing users buy is unconscious intervention. Autopilot adjusts temperature dozens of times per night based on real-time biometric data, moving through programmed thermal profiles calibrated to the user’s sleep stages without requiring any conscious input. The Pod 5 Ultra adds a further layer: when the system’s acoustic sensors detect the signature of snoring, the base automatically elevates the head to open the airway. The user wakes up having been intervened upon without being aware of it.

The third is frictionless data synthesis. The Pod tracks heart rate, HRV, and respiratory rate every night, building a longitudinal picture of physiological state without requiring the user to wear anything. For users who find wearables disruptive to sleep (or who simply don’t want yet another subscription device) the passivity of data capture is a meaningful differentiator.

Reviewers consistently describe the same experience: sleeping in a hotel after adapting to the Pod is miserable. The contrast with a static foam mattress, once experienced, is difficult to un-experience. Physical dependence of this kind is the strongest retention mechanic in consumer hardware. When the absence of the product causes physical discomfort, you know you’re onto something.

Retention Carries the Entire Model

Eight Sleep reached free cash flow positivity in 2025, a milestone tied to its refusal to compete on price. Gross margins are strong, and the company reports positive contribution margins on the hardware itself, before factoring in subscription revenue. This means the payback period on customer acquisition cost is effectively zero at point of sale, with every month generating subsequent high-margin subscription revenue against no incremental cost.

That model only holds if retention holds. Thirty percent of Eight Sleep’s revenue now comes from word-of-mouth referrals — a figure that indicates both genuine product satisfaction and an identity-driven evangelism among users who have made the Pod part of how they describe themselves. This organic flywheel lowers blended customer acquisition costs materially over time.

Structural churn risks exist in both directions. Earlier hardware generations suffered from water leaks — a failure mode that destroyed consumer trust and generated documented complaints about hydrogen peroxide residue forming under the cover. App redesigns have periodically forced users to relearn interfaces, introducing friction into a product that derives much of its value from invisibility. However, the dynamic most capable of undermining the model at scale is hardware decay. If the water pump infrastructure fails before the subscription lifetime value has matured, the replacement cost economics deteriorate significantly. The longevity of the physical hardware is not yet a known quantity.

Competing Without Fighting the Foam Wars

The competitive landscape splits cleanly once you set aside the legacy mattress industry. Tempur-Sealy and Serta-Simmons compete on brand awareness, retail distribution, and passive material technology — none of which matters against a software-driven thermal regulation platform, so they’re not really running the same race. The more relevant competitors are companies such as Sleepme, whose Dock Pro system offers water-cooled solutions at lower price points, and Sleep Number, which sells adjustable firmness with basic temperature controls. More recent entrants include Freshbed and Orion who offer very similar propositions.

Wearable companies (Oura, WHOOP, Garmin,…) compete for a different layer entirely. They measure biology and recommend behaviour changes, but they cannot physically intervene while the user is unconscious. Eight Sleep doesn’t just observe the body’s state. It manipulates the environment in real time based on that observation, creating a feedback loop no existing wearable can replicate from the wrist or finger.

Eight Sleep made one clever decision early that shaped its competitive position. It built a cover, not just a mattress.

By decoupling its technology from the underlying sleep surface, Eight Sleep made its addressable market the entire installed base of existing beds. A customer does not need to discard a $5,000 mattress to adopt the Pod. They place a cover on top of what they already own. This eliminated the most significant barrier to purchase in a category where inertia is high and replacement cycles are long.

The incumbent response to this is structurally self-defeating. Tempur-Pedic’s value proposition is built on the proprietary feel of its memory foam — an identity so central to the brand that building a thin technology layer on top of competitors’ beds would require the company to declare, implicitly, that the foam is not the product. Serta and Simmons are embedded in physical retail distribution networks built around the furniture buying experience. A channel that is completely misaligned with the software-subscription economics Eight Sleep runs. To compete directly, these companies would need to build firmware engineering teams, manage cloud computing costs, train AI models, and sell software subscriptions. They would need to become something they are not.

The strategic logic extends to the attention economy. Eight Sleep is screenless. The Pod communicates nothing visible to the outside world and makes no demand on the user’s focus. It is infrastructure (running in the background, accumulating data, adjusting the environment) while the user is unconscious. In a landscape increasingly defined by devices competing for attention, this invisibility is a product decision with profound strategic implications.

What Might Keep Them Up

Besides the price, this all sounds great, but as is the case with any business there are genuine risks the bullish narrative tends to absorb without accounting for.

Sensor accuracy is contested terrain. The Pod’s embedded mattress sensors operate at a distance from the body and are susceptible to additional interference that wrist-based and ring-based wearables largely avoid. A partner rolling across the midline, a pet joining the bed, or any significant movement can confuse the biometric readings. Sleep stage data generated by the Pod frequently diverges from data captured by dedicated wearables such as the Oura Ring, which sits closer to the body’s arterial system. For a platform increasingly building its value proposition on the depth and reliability of its health data, measurement inconsistency at the sensing layer is a serious vulnerability.

The October 2025 AWS outage was a warning. Eight Sleep is selling hardware to users who then become dependent on the company’s continued operation for that hardware to function at all. Customer support has confirmed that if the company ceased operations entirely, the hardware would stop functioning. The outage mode patch addresses the symptom. It doesn’t resolve the deeper tension: consumer hardware priced above $2,000 that depends on a single company’s servers to remain operable is a different kind of commitment than a monthly software subscription, and as the installed base grows, that tension will resurface.

The most significant risk is regulatory. Eight Sleep has filed with the FDA for clearance covering sleep apnea detection and mitigation — a move that signals the company’s ambition to cross from consumer wellness into clinical diagnostics. The company is also exploring the use of its data network to test the efficacy of supplements. Both directions lead into regulated territory where the rules are materially different from consumer hardware. The timelines are longer, the standards are higher, and the cost of a product failure shifts from a support ticket to a liability event. WHOOP offers an instructive case study in just how uncertain that path can be. In July 2025, the FDA sent WHOOP a warning letter over its Blood Pressure Insights feature, arguing it functioned as an unapproved medical device. WHOOP refused to remove it, and the two sides remained at a standoff for nearly a year before the FDA reversed course this June, closing its enforcement action after WHOOP made cosmetic changes to the feature and the agency issued new wellness-device guidance in the company’s favour. The dispute ultimately resolved well for WHOOP, but the lesson for Eight Sleep isn’t that regulatory risk is overstated. It’s that the line between a consumer wellness feature and a regulated medical device is genuinely unsettled, and a company pursuing FDA clearance should expect a prolonged, unpredictable process rather than a straightforward approval.

The Test Ahead

The Eight Sleep story is in many ways just getting started. They have built the hardware, validated the science, and seeded the cultural reframe. For the wealthy. It will take time before prices come down enough to make Eight Sleep products accessible to the average consumer.

Beyond that, the question remains whether a consumer hardware company can accumulate enough proprietary biological data, at sufficient scale, to build AI that outperforms episodic clinical data in predicting and preventing disease. The jury is out.

“Sleep is the greatest legal performance-enhancing drug that most people are neglecting.” — Matthew Walker

Roger Federer slept twelve hours a day during the tennis season. LeBron James has consistently prioritised ten. Usain Bolt structured his entire daily schedule around sleep windows. These figures reflect the extreme recovery demands of elite athletic training — the hours are higher than most people need because the physical load demands it. But the principle scales. These are data points from some of the most optimised performance environments on the planet, where every variable is scrutinised, every marginal gain competed for, and sleep keeps appearing at the top of the stack.

The lesson is not that founders should sleep twelve hours. It is that the highest-performing systems in the world treat sleep as a performance input, not a recovery afterthought.

The most effective performance intervention available to any athlete, founder, or high-performer costs nothing, requires no prescription, and is almost universally undervalued.

Sleep.

Recovery is invisible until it’s not

In elite sport, sleep is not framed as wellness or self-care. It is programmed. It sits underneath strength work, tactical preparation, skill development, and psychology — treated as a performance variable with direct impact on output, not an optional supplement to the real work. In startups, by contrast, sleep is routinely relegated to lifestyle advice. Something you optimise once the important decisions have been made. That mismatch is worth examining, because both environments demand the same thing from the people inside them: Sustained high performance under pressure.

We tend to romanticise intensity. Early mornings, late nights, the grind, the 996 hustle culture. Hard work is obviously integral to building anything extraordinary and lasting. Biologically and cognitively, however, intensity only pays off if the system is given space to adapt. You do not get fitter during training; you get fitter after training. You do not become sharper during the sprint; you become sharper between sprints.

That’s the trap with sleep loss. It doesn’t just make you slower — it makes you less aware of how compromised you are. A landmark 2003 study by Van Dongen et al. in the journal Sleep found that restricting sleep to six hours per night for two weeks produced cognitive deficits equivalent to two full nights of total sleep deprivation, and crucially… subjects consistently underestimated their own impairment. The number itself matters less than most people assume; quality and architecture matter as much as duration, and six hours of consolidated sleep can outperform eight hours of fragmented rest. But the miscalibration is the real culprit. You keep pushing because you feel functional, even as your decision-making quietly degrades.

It’s also worth distinguishing what sleep is doing. Slow-wave deep sleep is where physical repair and memory consolidation happen (i.e. where the body processes the load of the day and earns the right to perform again). REM sleep is where the cognitive work gets done: emotional regulation, creative pattern recognition, and the ability to connect disparate ideas are consolidated during REM. For founders, that second function is often the bottleneck. The functions sleep deprivation most visibly degrades — decision quality, creative thinking, and emotional regulation under pressure — are precisely what REM sleep exists to restore. Without it, you’re tired and operating with a structurally different cognitive profile.

In high-performance environments, effort is rarely the constraint. Capacity is.

The question is never just “Did you work hard?”

It’s “Are you ready to go again?”

Founders face the same equation, just with a different injury profile. They pull attention instead of hamstrings, accumulate cognitive fatigue rather than muscular damage, and the failure mode is subtle. Pattern recognition dulls, risk calibration shifts, patience shortens. Nothing collapses overnight, but leadership just becomes a little less sharp, a little less considered, and a little more reactive — until that becomes the new baseline.

The bottom line is that sleep is not a perk, but operational infrastructure — one that determines the quality of decisions, the tone of leadership, and the sustainability of intensity. Undermine it consistently, and you are borrowing performance from the future at compounding interest.

Technology is catching up

The category is moving from passive measurement toward active intervention. From generic dashboards to tighter feedback loops that facilitate actual behaviour change. The global sleep tech market was valued at around $25 billion in 2024 and is projected to exceed $100 billion by 2032, growing at close to 18% annually — one of the fastest growth rates in consumer health. That trajectory is driven by a fundamental shift in how people are beginning to think about sleep. As a powerful variable to be optimised for a better, longer, and high-functioning life.

The insight driving the category is that most performance errors are not caused by ignorance, but by miscalibration. A strong recovery tool therefore does not make you more intense. It acts as a 24/7 personal assistant, making you more aware, and helping you implement the right habits.

Wearables such as Oura and Ultrahuman are good examples of this. Their core contribution is the calibration of the vast amounts of data they collect. They take an internal, subjective state (”I feel fine”) and overlay it with objective constraint (”Your recovery is suppressed”). More importantly, these platforms are embedding sleep inside a holistic behavioural context. Sleep sits downstream from decisions made throughout the day — meal timing, glucose volatility, alcohol intake, stress load, training intensity. By tightening the causal loop between behaviour and consequence, the conversation shifts from “How did I sleep?” to “What did yesterday do to my sleep?” That personalised approach accelerates learning in a way that generic advice can’t.

Eight Sleep reframes the problem at a more fundamental level. Much of sleep optimisation has been positioned as personal discipline. Wind down earlier, reduce screen time, manage light exposure. Eight Sleep instead changes the environment directly by heating or cooling the bed, and on advanced models even changing physical elevation to reduce snoring. Temperature regulation for instance turns out to be one of the highest-signal variables in sleep quality, with several studies finding that optimised thermal conditions significantly improved deep sleep. Rather than demanding more willpower, Eight Sleep alters the default conditions. It treats sleep not as a discipline problem but as an environment design problem. The system changes first and the behaviour change follows.

Looking beyond how much you sleep and in what conditions, a third variable these platforms are beginning to surface is circadian alignment — when you sleep relative to your biology. Chronotype varies meaningfully between individuals, and founders operating across time zones or with irregular schedules can accumulate significant circadian debt even while logging adequate hours. You can bank eight hours at the wrong point in your circadian cycle and still impair the cognitive functions you most depend on. Timing is the dimension most people never measure, and increasingly it is what the better platforms are beginning to address.

So what?

The broader takeaway is not that every founder should buy a wearable or a smart bed. Eight Sleep’s price point makes it an option for the few, not the many, and it reinforces a broader problem around access in the longevity space that deserves its own piece. The lesson is architectural. Performance is built on intensity, but not on intensity alone. Recoverability needs to be part of the equation. The goal — whether in sport or in company-building — is to design conditions where performance compounds rather than oscillates.

For startups, this means treating recovery as a serious performance variable rather than a lifestyle footnote. It means designing for average energy and low motivation, not peak adrenaline — because a system that only works on your best day is poorly engineered. It means recognising that decision quality, consistency, and clarity compound over time in ways that occasional heroic output never does. McKinsey research has found that the four leadership behaviours most associated with high-performing executive teams — results orientation, problem-solving, seeking diverse perspectives, and supporting others — all rely heavily on the prefrontal cortex, the region of the brain most severely compromised by sleep deprivation. The cost of poor sleep is not just individual impairment; it is organisational drag. And it means changing environmental defaults before reaching for discipline, because defaults scale in a way that willpower never will.

But the harder realisation is that the barrier is rarely ignorance. Most founders know sleep matters. The real obstacle is culture. Startup environments actively signal that sleeping less is a performance indicator, and that signal is rarely explicit enough to argue with directly. It lives in who sends the 2am Slack message, in what gets celebrated in all-hands, in what leadership visibly models. Augmenting the default requires changing what the organisation demonstrates, not just what it says.

Most of us tend to worship visible output. Truly elite systems are built around invisible recovery.

Most wearables tell you what you’ve already done. Step counts. Calories burned. Workouts logged. They are, in essence, digital rearview mirrors — useful, perhaps, but fundamentally oriented toward the past.

WHOOP was built on the opposite bet: that the most valuable thing a device could do is tell you something you don’t already know. While humans are genuinely good at sensing their own physiological state, we still get certain things wrong. We confuse adrenaline with readiness. We mistake exhaustion for weakness. We train hard on days when we should be recovering, and call it discipline.

That insight began, in 2011, with a Harvard squash captain who noticed something elementary: athletes had no reliable way to know when they were recovered enough to train hard again. They went by feel. Will Ahmed (founder & CEO of WHOOP) spent the next several years and over 500 medical papers proving there was a better way — that Heart Rate Variability (HRV), the subtle variation in timing between heartbeats, was a precise and continuous window into whether the body was ready to perform or quietly breaking down. The problem at the time was that measuring it clinically required a $20,000 ECG machine and a hospital visit.

Ahmed set out to change that.

In a wearables market projected to reach ~$200 billion by 2030, and littered with the wreckage of hardware startups that failed to compete with Apple, Google, and Amazon, WHOOP has carved out a $10.1 billion valuation and a user base that borders on evangelical. What follows is an account of how, and a look at what comes next.

Technology Risk, Not Just Market Risk

WHOOP’s V1 prototype (aka Bobo) was not a sleek wearable. It was a bulky orange box tethered to a wall and a computer, built for a single purpose: to prove that clinical-grade HRV measurement from the wrist was even possible.

This is worth dwelling on. Most consumer hardware startups take on market risk — they build something using existing technology and bet that people will want it. WHOOP took on technology risk first. They spent years and millions of dollars solving a genuinely hard scientific problem before they had a product worth selling. The sensor had to be accurate enough to replace an ECG. The algorithms had to be validated against clinical standards. There were no shortcuts, because the entire value proposition depended on the data being trustworthy.

That upfront investment in technological credibility became the foundation of everything else. When WHOOP eventually told elite athletes that their device was as accurate as hospital equipment, it was true — and the athletes knew it, because the data matched what their own coaches and physiologists were seeing.

The depth of that investment also produced something nobody designed for. During the COVID-19 pandemic, WHOOP’s data science team noticed that a sudden spike in respiratory rate was a reliable leading indicator of infection — alerting users, including several PGA Tour golfers, to test themselves days before symptoms appeared. The device wasn’t built to do that. It just happened to be measuring the right things, continuously, at clinical resolution. That predictive capability — sensing what the body is doing before the conscious mind catches up — became WHOOP’s clearest proof of concept for everything that followed.

You can’t retrofit credibility.

Why Elite Athletes Actually Mattered

The conventional reading of WHOOP’s early strategy is that they used celebrity athletes as marketing. That somewhat misses the deeper logic. Targeting LeBron James, Michael Phelps, and Cristiano Ronaldo early served two purposes that had nothing to do with brand awareness.

The first was accuracy. Working with the world’s best athletes created a brutal feedback environment. These were people whose livelihoods depended on understanding their bodies with precision, supported by teams of physiologists and coaches who would immediately flag anything that didn’t hold up. If WHOOP’s data was wrong, they would know — and they would say so. This forced WHOOP to build the most accurate technology possible, preventing any temptation to cut corners for the mass market.

The second was cultural reframing. Historically, medical monitoring devices carried a stigma. If you wore one, something was wrong with you. WHOOP needed to invert that association entirely — to make health tracking feel not like a medical necessity but like an elite performance tool. Having it on the wrist of Olympians accomplished that in a way no advertising campaign could. When the NBA briefly banned players from wearing WHOOP on court and players started sneaking them on anyway, that told a story about desirability that money couldn’t buy. More recently, Sinner, Sabalenka and Alcaraz were banned from using fitness trackers, after the latter was busted for wearing a WHOOP under his wristband at the Australian Open.
Marketing stunt or not… the point stands: WHOOP has become, in the culture of elite sport, a signal that you are serious. Beyond the world of professional athletes it has become a mainstream status symbol for entrepreneurs, tech bros, VCs, and more broadly anyone who cares about their performance.

That cultural repositioning — from “medical device” to “high-status performance tool” — is one of the most important moves in the company’s history. The lesson isn’t to find famous ambassadors. It’s to identify the cultural frame that makes your product mean something different, and find the people who can make that frame credible. The awareness follows. It doesn’t lead.

From Hardware Sale to Health Subscription

WHOOP’s original commercial launch charged $500 upfront for the hardware. However, watching the valuations being attached to SaaS businesses, Will Ahmed made the contrarian decision to flip the model entirely: give the hardware away for free, and build the business on recurring subscription revenue.

The current packaging reflects a deliberate strategy to lower the barrier to entry while capturing increasing value from engaged users over time. The core tiers run from £169/$199 per year for the standard WHOOP 5.0, up to £349/$359 per year for the “WHOOP Life” tier, which unlocks medical-grade features including on-demand AFib detection and daily blood pressure monitoring. Critically, subscribers who maintain at least 12 months of remaining commitment receive hardware upgrades at no additional cost. It’s a powerful retention mechanic that makes the question of whether to upgrade your device disappear (kind of). Despite the company’s loyal customer base, the fine print caused Whoopgate — some intense online backlash after founding members and early adopters who were promised “free upgrades” still had to pay extra, when the WHOOP 5.0 came out last year. More on that below.

Beyond the core subscription, WHOOP has built a deliberate ecosystem of adjacent revenues with each layer expanding either the product’s surface area or its share of the user’s health spending. WHOOP Pro adds a $12 monthly fee for apparel discounts. WHOOP Body (smart garments that house the sensor in clothing rather than on the wrist) reinforces the “invisible infrastructure” positioning while opening a new hardware category. And straight from the Apple Watch playbook, premium accessory bands, including Italian leather options, address the fashion dimension without compromising the core product identity.

The more strategically significant moves are at the platform level. The integration of clinical blood test panels is an early land-grab on the user’s broader health spending — pulling laboratory diagnostics into the same subscription relationship as continuous monitoring. The women’s health expansion does something similar for an underserved market. Cycle phase has measurable effects on HRV, recovery capacity, and sleep quality that a dedicated period-tracking app cannot capture with the same fidelity. WHOOP’s continuous architecture is structurally well-suited to it in a way most competitors aren’t. For a platform historically associated with male-dominated performance sport, it is both a meaningful product differentiation play and a significant TAM expansion.

What Customers Are Actually Buying

When someone signs up for WHOOP, they are not purchasing a step-counter or a piece of fabric and plastic. They are entering into a continuous conversation with their own physiology. The value exchange operates across several distinct dimensions.

Predictive coaching. By quantifying both Strain (how much stress the body has accumulated) and Recovery (how well it has responded), WHOOP can tell users not just what happened yesterday, but what tonight’s sleep needs to look like to perform well tomorrow. This forward-looking orientation — living a step ahead of the user rather than behind them — is what sets it apart in the category.

Early-warning signals. The COVID case is the most dramatic example, but the principle applies broadly. WHOOP transitions health monitoring from a reactive system into a proactive one. Users report being alerted to overtraining, illness onset, and menstrual cycle disruptions before they consciously registered anything was wrong. That peace of mind, for the right user, is worth far more than the subscription price.

Identity and belonging. WHOOP is screenless, invisible under a sleeve, and communicates nothing to the outside world except that you wear it. Yet it has become a high-status signal in performance-oriented communities — a marker that you take your body seriously. That identity dimension drives organic word-of-mouth in a way that product features alone cannot.

Team and enterprise performance. Through WHOOP Unite, the product scales beyond the individual. Sports franchises, Formula 1 teams, the military, and corporations are using aggregated workforce data to manage recovery, reduce burnout, and optimise collective performance. This B2B layer represents a significant and still-underappreciated part of the commercial story.

Unit Economics and the J-Curve

Because WHOOP absorbs the hardware cost upfront, the cashflow shape is unusual and worth understanding clearly.

The early devices cost somewhere between $150 and $300 to manufacture. Over time, WHOOP has brought that down to an estimated $20-$50 depending on generation and volume (excl. R&D costs). WHOOP effectively finances that cost on behalf of the subscriber. With a $239 annual subscription (which is ideally for them paid upfront), the company is aiming to recover blended customer acquisition costs and hardware costs well within the first year. Every month a subscriber remains beyond that payback period generates high-margin software revenue with no incremental hardware cost. This is why retention is not just important to WHOOP — it is the entire model.

Customer acquisition costs are kept relatively low by the top-down flywheel that the elite athlete strategy created. Ronaldo, Sabalenka, Mahomes, the PGA Tour, and more recently organisations such as Ferrari F1 and PSG repping WHOOP in public generates aspiration and awareness that would cost hundreds of millions more in paid ads to replicate. The company has built a compelling brand that over time pulls customers toward it.

Annual churn (~10%), historically, has been lower than you might expect for a hardware subscription, with CAC/LTV reportedly sitting somewhere around the 4x mark. The reason is structural: WHOOP is one of the few wearables that demonstrably changes behaviour over time. Users who stay on the platform for a year measurably lower their resting heart rate, improve sleep consistency, and develop new habits around recovery. That said, a growing number of consumers stop paying the monthly subscription and churn once they have understood their core health patterns — a dynamic that Advanced Labs is directly designed to address, by continuously raising the ceiling on what the platform can tell you about yourself.

The Health OS Race

WHOOP Advanced Labs, launched in September 2025, is the clearest signal yet of where Will Ahmed is actually trying to go. The feature integrates clinical blood testing directly into the WHOOP platform — covering 65 biomarkers across hormones, heart health, metabolic function, inflammation, and cognitive performance, with clinician-reviewed results and AI coaching delivered inside the app, powered by Quest Diagnostics. Over 350,000 members had joined the waitlist before the feature even launched — a level of pent-up demand that tells you something important about what WHOOP’s user base actually wants.

The surface-level read is that this is a smart revenue extension — a high-margin add-on that deepens engagement and raises the subscription ceiling. That’s true. But it undersells the strategic logic. What WHOOP is building toward is something more ambitious: the longitudinal health record that sits at the centre of a person’s entire health operating system (OS). Continuous biometric data from the wrist, periodic bloodwork from a lab, clinical AI interpreting the relationship between the two, and coaching that updates in real time as your body changes. In 2026, that integration has deepened further — biomarker data now flows directly into each of WHOOP’s core Healthspan pillars, with AI providing contextual guidance on what the connections mean and what to do about them. The device is no longer just measuring you. It’s building a model of you.

This is the real competitive battleground — and WHOOP is not running it alone.

In 2025 and into 2026, every major player in the wearable health space crossed the same threshold: they stopped competing on hardware specs and started competing on intelligence — who has the best AI, the most contextual data, and the stickiest coaching loop. Oura overhauled its app, making its AI Advisor genuinely contextual by finally giving it access to personal health data. Garmin launched Connect+, a paid subscription tier layering AI coaching on top of its historically hardware-led model — a significant strategic pivot for a brand that built its reputation on not needing subscriptions. Apple, Samsung, Ultrahuman, and a wave of emerging challengers are all converging on the same layer. Simultaneously, companies such as Neko Health, Superpower, and Function Health which initially ran blood and imaging tests, are now vying for a piece of the Health OS pie. And at the foundation layer, Amazon, Anthropic, and OpenAI are building health intelligence capabilities that could eventually sit beneath all of them.

The winner of this race won’t be determined by sensor quality. It will be the platform that accumulates the deepest longitudinal data, builds the most personalised, actionable AI on top of it, and makes leaving feel genuinely costly — because leaving means abandoning years of your own health history, calibrated specifically to your physiology. On that dimension, WHOOP’s head start is real. The company is already training its AI models on over 24 billion hours of continuous physiological data — a dataset no competitor can replicate without years of similar scale and commitment. By combining that foundation with over 100,000 daily data points per user and blood biomarker data, WHOOP is making a credible case for being the most complete and actionable health platform available to consumers. Whether they can execute without stumbling into the regulatory friction that ambition inevitably attracts is the question the next section addresses.

Focus as a Competitive Strategy

WHOOP has never tried to out-build Apple. That would be a losing game. Instead, it has won by refusing to play Apple’s game at all.

The screenless design is the clearest expression of this. By eliminating the screen, WHOOP gave up notifications, apps, contactless payments, and everything else that makes the Apple Watch a Swiss Army knife. What it gained was longer battery life, a smaller form factor, an “invisible” aesthetic that doesn’t announce itself as a tech device, and complete freedom from the attention economy. WHOOP is never competing for your focus. Instead, it gives you focus. It is only ever in the background, monitoring, accumulating signal.

Data privacy is a second structural advantage. In a landscape where user data is routinely monetised — a practice that notoriously complicated Google’s acquisition of Fitbit and triggered regulatory scrutiny — WHOOP has committed explicitly to not selling customer data to third parties. For health data in particular, that commitment resonates with users in a way that general privacy policies do not.

The third element is a deliberately bifurcated innovation cycle. Hardware roadmaps run three to four years out, placing ambitious bets on unproven technology and giving engineering teams time to solve genuinely hard problems. Software ships on two-week sprint cycles, constantly delivering new value to existing subscribers and giving the team rapid feedback on what’s working. These two clocks serve different purposes and don’t need to be synchronised — a distinction that most hardware companies fail to make clearly.

None of this is replicable by Apple without Apple becoming something it isn’t. That’s the point. The most durable competitive positions aren’t the ones that are hard to attack — they’re the ones that are impossible to attack without self-destruction.

Where It Gets Harder

A clear-eyed read of WHOOP requires naming what the company gets wrong, or what might go wrong — because there are genuine risks that the bullish narrative tends to absorb without accounting for.

Sensor accuracy is still contested. The entire value proposition of Strain and Recovery depends on the heart rate data being reliable. Independent reviewers have found WHOOP’s HR readings inconsistent during exercise — sometimes too high, sometimes too low, and sometimes accurate — a pattern that is arguably more problematic than being consistently wrong, because it makes it harder to know when to trust the data. Oura’s ring-based sensors, sitting closer to the arteries of the finger, have demonstrated a structural signal quality advantage for certain metrics — one that WHOOP cannot easily engineer away from the wrist. For a platform increasingly making clinical-adjacent claims, the measurement layer underneath those claims still has meaningful limitations.

The upgrade backlash was a structural warning, not just a PR problem. The WHOOP 5.0 launch in May 2025 triggered significant consumer backlash — Reddit threads titled “Whoopgate: The Receipts”, Bloomberg coverage of WHOOP “trending for all the wrong reasons” — after the company appeared to contradict its own prior commitments about who qualified for free hardware upgrades. Long-term subscribers who had paid upwards of $800 over four years found themselves potentially ineligible for the upgrade that new customers received immediately upon joining. WHOOP reversed course and issued refunds, but trust issues linger, particularly around transparency. For a business model built entirely on retention economics and community loyalty, this kind of trust erosion matters beyond the PR cycle. The incident revealed a tension that will resurface: as WHOOP segments its product tiers upward and adds premium-only features, it risks alienating the loyal base that built the brand.

The FDA is watching the medical device boundary closely. This is the most significant structural risk and the one most likely to shape WHOOP’s trajectory over the next several years. In July 2025, the FDA issued a warning letter regarding WHOOP’s Blood Pressure Insights feature, concluding it constitutes a medical device based on intended use — estimating blood pressure — and therefore requires premarket authorisation WHOOP had not obtained. That letter anchored a class action lawsuit filed in November 2025, in which plaintiffs alleged consumers paid a premium for features that were not legally marketed, seeking restitution and an injunction against further promotion of the blood pressure functionality. The case is unresolved and the allegations remain unproven. But the underlying tension is very present: the more WHOOP pushes toward clinical-grade claims to justify premium pricing, the more it enters regulated territory — where the rules are different, the timelines are long, and the cost of getting it wrong is material. Becoming a health OS and remaining a consumer wearable are not inherently incompatible, but the path between them runs directly through regulatory approval processes that are slow, uncertain, and expensive.

None of this diminishes what WHOOP has built. But it does reframe the question for anyone thinking about this space seriously: the company’s next chapter is less about whether the platform works, and more about whether it can navigate the transition from performance tool to health infrastructure without losing the trust, focus, and regulatory clarity that got it here.

Conclusion

The WHOOP story is unusually interesting precisely because it is still ongoing. The first chapter — building the technology, seeding the culture, structuring the model — is largely settled. The second chapter is just getting started.

What the next phase tests is whether those same principles hold as WHOOP pushes deeper into clinical territory, where the regulatory environment is less forgiving, the competitive field is broader, and the definition of “performance” is expanding faster than any single company can contain.

"You do not rise to the level of your goals. You fall to the level of your systems"
— James Clear

Phil Jackson won 11 NBA championships. He also burned sage in locker rooms, made players read Zen philosophy, and let Dennis Rodman fly to Las Vegas to wrestle mid-season (a special case, I admit). None of that is in any coaching manual, but all of it was deliberate. Jackson wasn’t managing behaviour, he was building rituals that made the mundane sacred and kept ego out of the system.

He’s a good place to start. Because the most durable performers in sports share three things that have nothing to do with talent or tactics: they design environments, not features. They build rituals, not goals. And they measure what compounds, not what flatters.

Design the system, not the feature

When Pep Guardiola took over Manchester City, he didn’t start with tactics. He started with details that seemed minute at the time: how players interacted with each other, how the shape of the dressing room impacted culture, and how video analysis and data were used in the daily routine. No new formation or marquee signing (until later). He first redesigned the environment itself, and elite performance followed.

James Vowles is doing the same thing at Williams Racing, just with higher stakes and a longer timeline. He took over with an explicit mandate to “break everything”. Legacy systems, outdated processes, a culture built on past glory rather than present reality — he’s dismantling all of it, deliberately and without sentiment. He’s implementing new technology stacks, new accountability structures, new ways of working at every level. The early results have been a mixed bag, for sure. But Vowles isn’t building for this season. He’s rebuilding the organisation as a system, and that requires completely breaking it first. That’s a fundamentally different game. Much slower and harder, but the only one worth playing if you want sustained performance rather than a lucky podium.

Both cases tell the same story: durable performance is never the product of a single intervention. It’s what happens when the environment itself is designed to produce it. For a founder, that means asking a harder question than “does this feature work?” It means asking: does our product change the conditions under which the user operates — or does it just add another variable to an underlying system we haven’t touched?

Rituals beat goals

Goals tell people where to go. Rituals get them there, and crucially keep them there when motivation runs out.

Coach Jackson from the example above understood this intuitively, and he built it into the architecture of his teams. His famous Triangle Offense wasn’t just a set of plays drawn on a whiteboard — it was a philosophy encoded as a system, designed to turn five individuals into a single organism reading and reacting in real time. No one player could dominate it. No one player could break it. The system made the collective smarter than any individual within it. Off the court, he reinforced the same logic through ritual: burning sage before games, assigning philosophy texts, giving Rodman permission to miss practice and wrestle in Vegas — because he understood that conformity kills creativity, and that a system capable of holding Dennis Rodman is a system capable of holding anyone. When egos clashed, he benched them. Not the players, the egos. The Triangle didn’t care who you were. It only cared whether you played your role within it.

The New Zealand All Blacks built something similar, and they put it in simpler terms: “Sweep the sheds.” After every match, all players (including the most senior) clean the locker room. This is not about hygiene. It’s about ego. A signal, repeated until it becomes reflex, that no one on the team is above the work. No task is ever beneath you. Over time, the behaviour became so embedded it stopped needing enforcement. It just was what the All Blacks did.

For a product builder, the implication is direct: make the key behaviour the default. Give it a clear cue, a cadence, and a defined recovery path for when someone slips. A goal tells your user what they should do. A ritual is what your product does when they don’t feel like it.

Measure what compounds, not what flatters

Sir Dave Brailsford built British Cycling around a deceptively simple idea he called the aggregation of marginal gains: find every variable that affects performance, improve each one by 1%, and let the compounding do the work. Redesign the bike seat. Test fabrics in wind tunnels. Teach riders the optimal way to wash their hands to reduce illness. Individually, none of it moves the needle. Collectively, it moved the entire sport. Critics thought he was insane. Then British Cycling dominated the Olympics and the Tour de France. For a decade.

The insight wasn’t that any single optimisation mattered. It was that the critics were measuring the wrong things — evaluating each intervention in isolation, asking whether this or that change made a difference. Brailsford was measuring what happened when everything compounded. The aggregate. The system-level output over time. That’s a completely different question, and it requires completely different metrics to answer it. The 1% gains only work if you’re tracking the sum, not the parts.

For companies building in performance and health, this cuts directly against how most products are evaluated. Biomechanics, load management, VO2 max — undoubtedly useful, but they measure the variable, not the system. Adherence, engagement, time-to-value are harder to report on, but actually tell you whether your product is working.

So what?

The leaders in these examples share one insight: you can’t isolate and optimise variables independently. Everything interacts. Sleep affects training. Training affects recovery. Recovery affects cognition. Cognition affects culture. Culture affects everything.

The companies that understand this — that build products treating humans as complex adaptive systems rather than machines to be tuned — will win. In the era of AI, that distinction matters more than ever.

The hardest problems in human performance aren’t technical. They’re environmental, ritual, and cumulative. The best founders aren’t solving for the motivated user on their best day. They’re building systems that hold on the average Tuesday, when no one feels like it and the habit still needs to hold.

At Momentous, we back founders who understand this, and build accordingly.