Signals Inbox·July 23, 2026·Data Centers

Is Fluidstack really worth $7.5B today?

Fluidstack’s $7.5 billion valuation is aggressive but plausible: Anthropic, more than 900 megawatts of capacity and unusually fast growth support the bet, while thin verified revenue, low historical margins and huge long-term commitments leave almost no room for delays.

Q1What actually happened to Fluidstack’s $7.5 billion valuation?

Fluidstack’s $7.5 billion valuation is now confirmed, although a newer SEC filing shows that another large equity raise is already under way.

Fluidstack has finally disclosed that it raised $830 million in a Series A led by Situational Awareness, the investment firm founded by former OpenAI researcher Leopold Aschenbrenner. The financing closed earlier in the year, several months before Fluidstack announced it publicly.

The size is extraordinary for a Series A. Fluidstack was incorporated in the United Kingdom in 2017 and had raised only a fraction of that amount before its recent infrastructure expansion. It reached $7.5 billion a little over eight years after being founded, roughly the same age at which CoreWeave entered the public market.

There is already a newer development. Fluidstack filed an SEC Form D for a separate $1.5 billion equity offering and reported approximately $730 million sold to seven investors. Morgan Stanley is acting as placement agent. The filing does not reveal the valuation attached to those shares.

For now, $7.5 billion remains Fluidstack’s last confirmed valuation. Reports of talks at an $18 billion price may eventually prove accurate, but the latest regulatory filing does not confirm that figure.

Q2How much revenue does Fluidstack really make today?

Fluidstack’s verified revenue is still far too small to support a $7.5 billion valuation on its own.

The company’s latest UK accounts show £52.7 million of revenue in 2024, equivalent to roughly $66 million at the exchange rate used in contemporary reporting. Revenue had been £21.9 million the previous year, so the business grew by around 140%.

Sacra estimates that Fluidstack finished 2024 with approximately $180 million in annual recurring revenue, up from an estimated $25 million one year earlier. Fluidstack has never confirmed that figure, and Sacra’s number measures the annualized value of active business rather than revenue already recognized in the accounts.

The gap between $66 million of recorded sales and $180 million of estimated ARR is possible for a company signing large contracts late in the year. Still, we should treat $180 million as a generous working estimate rather than a verified company metric.

Fluidstack’s 2025 accounts have yet to appear publicly. Until they do, nobody outside the company can see how much of its announced expansion has turned into real revenue.

Q3Is Fluidstack really valued at 42 times revenue?

On the most generous public estimate available, Fluidstack is currently priced at about 42 times ARR.

Dividing the $7.5 billion valuation by Sacra’s $180 million estimate gives a multiple of 41.7. Using Fluidstack’s audited 2024 revenue produces a multiple of roughly 114.

The second figure is clearly backward-looking because Fluidstack’s business has expanded since those accounts closed. The 42-times estimate already gives the company credit for that expansion, and it remains extremely high.

Investors usually reserve multiples around this level for software companies with high gross margins, limited equipment spending and revenue that can grow without building physical infrastructure. Fluidstack needs data centers, power contracts, networking equipment and accelerators costing billions of dollars.

At 42 times ARR, the valuation already assumes several more years of exceptional growth. Merely participating in a fast market will not be enough.

Q4Is Fluidstack still growing fast enough?

Fluidstack is growing at a rare pace, and that is the strongest financial argument for its valuation.

Audited revenue increased by approximately 140% in 2024. Sacra’s ARR estimates imply much faster growth, from $25 million to $180 million, although those figures remain unverified.

Either way, Fluidstack has clearly moved beyond its earlier marketplace model. Its business now includes private GPU clusters, managed AI infrastructure and large data-center projects designed for frontier laboratories.

Growth from a small base can look spectacular without producing a giant business. Another year of 140% growth from $66 million would produce roughly $158 million of annual revenue. Starting from the higher $180 million ARR estimate would give around $432 million.

That second outcome would begin to support the valuation. The first would leave Fluidstack extremely expensive. The next set of accounts will therefore carry far more weight than another infrastructure announcement.

Q5Does Anthropic’s $50 billion buildout prove Fluidstack is worth $7.5 billion?

Anthropic’s $50 billion infrastructure program makes Fluidstack strategically important today, yet it does not reveal how much revenue Fluidstack will actually keep.

Anthropic selected Fluidstack to build custom data centers in Texas and New York as part of a broader $50 billion investment in American computing infrastructure. The first facilities are expected to begin operating during 2026, with more sites planned afterward.

That relationship moved Fluidstack into a different category. Investors are now pricing it as a major infrastructure partner for frontier AI laboratories, responsible for power, construction, hardware deployment and operations.

The $50 billion covers Anthropic’s full infrastructure program. Fluidstack’s share has never been disclosed. We also do not know how much will go to GPUs, landlords, utilities, construction companies, lenders and other suppliers.

Anthropic’s commitment proves that Fluidstack can win projects at enormous scale. What it does not prove is whether those projects will produce the revenue and margins required by a $7.5 billion valuation.

Q6Are Fluidstack’s multibillion-dollar contracts actually revenue?

Fluidstack has secured more than 940 megawatts of capacity, while the widely quoted $26.2 billion belongs to its landlords’ revenue books.

TeraWulf, Cipher Mining and Hut 8 have announced four major agreements involving Fluidstack. Together, the initial contracts cover more than 941 megawatts of critical computing capacity and could generate approximately $26.2 billion for the companies providing the sites.

Fluidstack sits on the paying side of those contracts. It must place customer workloads into the facilities at prices that cover rent, power, hardware, financing and operations. The deals provide scarce capacity, but they also create long-term financial commitments.

Google has backstopped at least $5.9 billion of Fluidstack’s disclosed obligations across the TeraWulf and Cipher projects. That support makes the developments easier to finance and reduces the risk faced by the landlords.

The contracts show that Fluidstack can assemble infrastructure on a huge scale. They reveal very little about Fluidstack’s own future revenue.

Fluidstack-linked infrastructure agreements

Q7How expensive is Fluidstack next to CoreWeave and Nebius?

Fluidstack currently carries a clear premium to CoreWeave and Nebius, even after we give it credit for being smaller and earlier in its growth.

At recent market prices, CoreWeave was worth around $43 billion. It generated $2.08 billion of revenue in the first quarter, more than double the previous year’s figure. Annualizing that quarter gives approximately $8.3 billion, placing its equity value near five times revenue.

CoreWeave also had $99.4 billion of backlog and more than one gigawatt of active power. Backlog and active power are not revenue, but they show a business already operating at a scale Fluidstack is still trying to reach.

Nebius was recently worth around $55 billion. It reported $1.9 billion of AI cloud ARR and guided to between $3 billion and $3.4 billion of full-year revenue. Its market value equals roughly 29 times AI ARR or 17 times the midpoint of its revenue forecast.

Fluidstack may grow faster because it starts from a much smaller base. Even so, the gap is too large to dismiss as a normal private-market premium. Investors are paying Fluidstack more per dollar of estimated revenue than they pay for two public companies already reporting their numbers every quarter.

Fluidstack compared with CoreWeave and Nebius

Q8Is Fluidstack pricier than private neocloud rivals?

Fluidstack also looks expensive beside the closest private competitors with usable revenue estimates.

Together AI recently raised $800 million at an $8.3 billion valuation. The company said annual bookings had passed $1.15 billion, while Sacra estimated approximately $1 billion of annualized revenue. That would place Together near eight times estimated revenue.

Lambda’s most recent primary valuation was approximately $5.9 billion. Forge’s secondary-market data later suggested a value closer to $9 billion. Against Sacra’s estimated $760 million revenue run rate, Lambda sits between roughly eight and 12 times revenue.

Together and Lambda are not exact twins. Together sells a broader AI platform, while Lambda combines cloud infrastructure with hardware and financing. Their revenue estimates also come from outside the companies.

The difference is still too large to ignore. Fluidstack’s estimated multiple is three to five times higher. That is a big gap, even in this market.

Nscale provides the more aggressive precedent. It recently raised $2 billion at a $14.6 billion valuation despite offering little public financial information. That round shows how strongly private investors currently value power access and future data-center capacity. Without revenue, it cannot tell us whether Fluidstack’s own multiple is sensible.

Q9Is AI infrastructure demand still strong today?

AI infrastructure demand is still expanding quickly enough to support several giant companies.

Synergy Research estimates that neocloud providers generated more than $25 billion of revenue in 2025. Fourth-quarter revenue alone reached about $9 billion, rising 223% from a year earlier. Synergy expects the market to approach $400 billion annually by 2031.

IDC recently measured $89.7 billion of worldwide AI infrastructure spending in one quarter, up 33% from the previous year. Spending was roughly flat from the preceding quarter, so the market is growing rapidly without accelerating every three months.

There is plenty of money moving into this industry. The harder question is how much Fluidstack keeps after paying for chips, power, rent, construction and financing.

Competition is also broadening. Meta has reportedly explored selling excess AI computing capacity and discussed a possible multibillion-dollar arrangement with Anthropic. If companies building models also become infrastructure suppliers, customers will have more options and GPU prices could come under pressure.

Fluidstack has chosen the right market. That alone cannot protect its valuation.

Q10What can Fluidstack do that competitors cannot easily copy?

Fluidstack’s strongest advantage today is execution speed: it can assemble power, buildings, financing and GPUs faster than most customers could manage alone.

Fluidstack says it can bring gigawatt-scale projects online in around six months, compared with an industry norm of 18 to 24 months. That comparison comes from the company, so we cannot treat it as an independent benchmark. Its work for Poolside provides a more concrete example: Fluidstack deployed more than 2,500 GPUs within 48 hours.

The product also includes managed Kubernetes and Slurm clusters, high-performance networking and scheduling software designed around the physical layout of the GPUs. Customers receive an operating cluster instead of a collection of individual machines.

That combination saves AI laboratories from negotiating separately with chip suppliers, data-center operators, utilities, networking vendors and lenders. For a frontier laboratory racing to train its next model, several months can be worth far more than a small difference in hourly GPU prices.

Competitors can copy parts of this model. Reproducing the full operation takes power contracts, engineers, construction expertise, financing relationships and credibility with large customers. Fluidstack has a real head start. Keeping it is the hard part.

Q11Is Fluidstack becoming an owner, or can customers cut it out?

Fluidstack is moving closer to asset ownership, while Anthropic is also showing that it can contract directly with data-center operators.

A recent TeraWulf transaction captures both sides of the story. TeraWulf agreed to sell its 50.1% interest in the 168-megawatt Abernathy joint venture to an investor group led by Fluidstack. That gives Fluidstack greater control over a major site and potentially a larger share of its long-term economics.

At the same time, TeraWulf signed a direct 20-year lease with Anthropic for a separate 401-megawatt campus in Kentucky. TeraWulf expects that agreement to generate roughly $19 billion of contracted lease revenue, with the first phase scheduled for the second half of 2027.

Fluidstack therefore has a credible path toward owning more of the infrastructure it develops. Direct deals between Anthropic and landlords show that Fluidstack will not automatically sit inside every future project.

Its long-term value will depend on whether customers see Fluidstack as essential to designing and operating these clusters, rather than as an intermediary they can eventually bypass.

Q12Does Google’s backing strengthen Fluidstack or expose a dependency?

Google’s backing materially strengthens Fluidstack, and it also reveals how much the expansion relies on a much larger balance sheet.

Google has guaranteed at least $5.9 billion of Fluidstack-related obligations across disclosed TeraWulf and Cipher projects. Those guarantees allow landlords to finance construction using Google’s credit quality instead of relying entirely on Fluidstack’s limited financial history.

The trade is straightforward. Fluidstack gains access to projects that would be difficult to finance alone. Google gets equity exposure to infrastructure partners and another route for placing its TPU accelerators outside Google Cloud.

Reports have also linked Google to a possible direct investment of around $100 million in Fluidstack. The company has yet to confirm that transaction.

The partnership is valuable. Its durability depends on Fluidstack continuing to serve Google’s infrastructure and TPU strategy better than other providers could.

Q13Can Fluidstack’s margins support a $7.5 billion valuation?

Fluidstack’s latest audited margins are nowhere near strong enough for the revenue multiple shown above.

The company generated £52.7 million of revenue in 2024. Gross profit was approximately £6 million, giving Fluidstack a gross margin of only 11.4%. Operating profit was £666,000, and the group finished the year with a £556,000 net loss.

An 11% gross margin leaves little room for salaries, financing costs, administration and unexpected project delays. Fluidstack also spent more than £44 million on computing equipment during the year, nearly matching its annual sales.

The newer data-center model could produce better economics. Larger customers may sign longer contracts, hardware utilization may improve and Fluidstack may retain more value as it moves closer to ownership.

We do not yet have financial statements showing that improvement. Investors have assumed the business mix and margins have changed much faster than the published accounts.

Q14Is Fluidstack too dependent on Anthropic and a few large customers?

Fluidstack is currently too dependent on a small number of large customers, with Anthropic sitting at the center of its biggest expansion.

Fluidstack itself identified customer concentration as a risk in its 2024 strategic report. Large contracts helped the company grow quickly, but they also give each customer considerable negotiating power.

The problem becomes more serious when Fluidstack signs long leases to serve those customers. A frontier laboratory can delay a facility, change its hardware plans or shift future workloads elsewhere. Fluidstack may still owe money to landlords and lenders.

Long contracts can reduce this risk when customer commitments closely match Fluidstack’s own obligations. Public announcements rarely reveal termination rights, pricing adjustments, utilization guarantees or responsibility for unused capacity.

Anthropic could transform Fluidstack. It also makes one company’s capital plans unusually important to Fluidstack’s performance.

Q15How much revenue would Fluidstack need to justify $7.5 billion?

Fluidstack needs roughly $375 million to $500 million of annual revenue for its valuation to look demanding rather than extreme.

At $500 million, the $7.5 billion valuation would equal 15 times revenue. At $375 million, it would equal 20 times. Both would remain expensive, although recent public-market valuations for fast-growing AI infrastructure companies show that investors may accept those levels.

Using the $180 million working estimate discussed above, Fluidstack must more than double revenue to reach $375 million. Getting to $500 million requires growth of approximately 178%.

A 10-times multiple would require $750 million of revenue. That level would place Fluidstack much closer to the valuation range of a scaled infrastructure provider.

Revenue required at different valuation multiples

Q16Can Fluidstack grow into the valuation within two years?

Fluidstack can grow into $7.5 billion within two years, but it needs another stretch of exceptional execution.

If revenue doubled in each of the next two years, the estimated $180 million base would become $720 million. The valuation would then equal about 10 times revenue, a reasonable level for a company still growing rapidly.

Two years of 70% annual growth would produce roughly $520 million and a multiple near 14 times. That outcome could also support the price if margins improved.

At 50% annual growth, revenue would reach around $405 million and the valuation would remain above 18 times revenue. Investors would still be paying heavily for future expansion.

The timing is tight because capacity announcements arrive well before revenue. Several large facilities are scheduled to open between 2026 and 2028. Permitting, grid connections, construction delays and hardware deliveries could push revenue beyond the period investors currently appear to be pricing.

Q17What has to go right, and what would break the valuation?

The bull case requires four things to happen together: more revenue, better margins, timely openings and a broader customer mix.

Fluidstack needs its current revenue run rate to move toward $400 million or more, while keeping growth well above the broader infrastructure market. The Anthropic facilities must open close to schedule, customers must use the capacity they reserved, and the company must retain more than the 11% gross margin shown in its latest accounts.

Greater ownership could help. Fluidstack’s move to lead the Abernathy investor group may let it capture property and infrastructure economics that previously flowed to landlords. A larger portfolio of customers would also make the company less vulnerable to one laboratory changing its plans.

The bear case begins with delays. Long-term leases continue while customer revenue arrives later than expected. More capacity from CoreWeave, Nebius, Meta and the hyperscalers then pushes prices down. Fluidstack could end up carrying expensive commitments while stronger companies control the chips, customers and financing.

A slowdown would be particularly painful at this valuation. Investors have already paid for a large part of the future growth. There is not much slack here.

Q18So, is Fluidstack really worth $7.5 billion today?

Fluidstack is worth $7.5 billion only at the edge of what today’s evidence can support.

Our verdict is aggressive but plausible. Fluidstack has achieved something rare: it won a central role in Anthropic’s infrastructure expansion, secured access to more than 900 megawatts and raised enough capital to compete for projects once reserved for hyperscalers.

The financial evidence remains much weaker. Verified revenue is small, estimated ARR implies a very high multiple, historical margins are thin and the company relies heavily on a few customers and financial partners.

The $7.5 billion valuation makes sense if Fluidstack’s current revenue is already heading toward $400 million to $500 million, its larger contracts produce much better margins and its major facilities open without serious delays. The company has provided credible reasons to believe this could happen. It has yet to provide the financial results proving that it has happened.

The latest SEC filing suggests that investors may already be funding Fluidstack at a newer price. Since the filing omits the valuation, $7.5 billion is still the only figure we can properly judge.

At that price, we would not call Fluidstack cheap. We would call it a bold valuation attached to a company with a real chance of growing into it. Any move toward the reported $18 billion level would be far harder to defend with the numbers currently available.