Invest1 distinct publisher3 min readPublished
Two compute landlords raised $4.5bn between them while cloud security, protein bars, payroll software and robot silicon split under $900m, and the second-biggest cheque came from a trading firm rather than a venture fund.
The Investor · Invest desk

Compiled by The InvestorSomething wrong?How this is made
Price per dollar ever put in is the one comparison this table makes easy. Crusoe's $30bn mark sits on nearly $7.2bn raised across the company's life, about 4.2 times capital in [3][6]; Fluidstack's $18bn sits on just over $2.6bn, roughly 6.9 times [4][7], so the smaller company is carried some 66 percent richer per dollar it has ever taken [8]. Run the tripling backwards and the year gets clearer: $30bn over three is about $10bn, set against the $4.2bn raised before this round, or 2.4 times capital in [9]. The re-rating moved further than the fundraising did.
Multiple of capital in is also close to useless for a business that rents machines, because it counts only the equity, and the roundup discloses no contract durations, no capital expenditure plan, no debt and no depreciation life for either company [12]. The payback question that hangs over any data centre is not answered here; the more interesting version of it is not even posed, since $3bn buys about a tenth of Crusoe at the marked price [10] and $1.5bn about a twelfth of Fluidstack [11], with not one published lease term underneath either.
Crunchbase calls the Fluidstack round a private equity round led by Jane Street Capital [4], which is a proprietary trading firm writing the second-largest cheque of an American venture week [11], and $1.5bn is about 58 percent of every dollar the company has raised since inception [12]. Terms are not disclosed [12]. In a capacity business where the asset has a known purchase price and an unknown residual value, the preference structure is most of the investment, and we have the headline number without it.
Across the seven rounds this roundup details, the three AI infrastructure deals take $4.8bn of $5.68bn [2][4], and the top two alone take 79 percent [3]. Valor Equity Partners co-led both Crusoe's $3bn and David's $250m protein-food round in the same week [1][8], which is either a genuinely diversified book or one committee pricing two assets it has no common yardstick for.
This material leaves two readings open. The roundup names payments and healthcare among the week's ten but details only seven amounts [13], so the denominator is soft. And if equity is a minority of the capital stack at these two companies, then venture is not the marginal dollar at all, the lenders are, and a $3bn Series F is the thin visible slice of a much larger financing. The read that compute is absorbing the marginal dollar breaks the moment one of these companies publishes a stack showing debt at several times equity, because at that point the price that decides the outcome is the coupon rather than the mark.
Ranked by verification strength, evidence, and original report placement.
Denver-based Crusoe raised a $3 billion Series F co-led by Atreides Management and Valor Equity Partners, with Mubadala Capital also participating.
Crusoe, originally founded to use stranded natural gas to power cryptocurrency mining, has become an AI cloud and data center provider serving customers including OpenAI, Microsoft and Meta.
Crusoe has raised nearly $7.2 billion to date, and the latest financing values it at $30 billion, triple its valuation of less than a year ago, according to Crunchbase.
New York-based Fluidstack raised $1.5 billion in a private equity round led by Jane Street Capital, bringing total funding to just over $2.6 billion and valuing the company at $18 billion.
Fluidstack provides large-scale GPU and data center infrastructure for demanding AI workloads and is one of a growing group of companies spending heavily to meet AI compute demand.
San Francisco-based Gimlet Labs raised a $300 million Series B led by Andreessen Horowitz, with Sapphire Ventures, Menlo Ventures, Arm Holdings and M12 among investors; it is building an AI inference cloud that distributes workloads across different types of chips, has raised $392 million to date and was valued at $3 billion.
Distinct publishers with included, body-backed reporting in this cluster.
Follow any of these and your For You feed starts watching them — no settings page required.
product
Rillet's $100M reads as proof mid-market ERP is rip-and-replace, mostly at the cheap end1 distinct publisher
product
Washington's secret AI test is coming for open weights, and release dates go with it2 distinct publishers
product
Meta owns the models and the data centres, and still pays Microsoft to rent someone else's1 distinct publisher
leadership
Meta scraps Project OT, underscoring how hard AI deployment really is1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One aggregator citing its own database
Every dollar figure reaches the reader through Crunchbase reporting Crunchbase data, with no filing, company statement or investor confirmation behind any of it. The internal arithmetic is sound and the round descriptions are specific about who led and who followed. What the reporting cannot establish is what $30bn and $18bn actually price, because the structure of the Jane Street-led round was never published and neither compute business's contracts, capex or debt appear.
Three customer names, zero terms
The only demand signal in the whole roundup is a list of three Crusoe customers. Fluidstack, which took $1.5bn, is credited with meeting soaring demand without a single customer being named. Money committed is not capacity taken up, and this reporting measures the former.
Four fifths needs a smaller denominator
The concentration is real, but our own headline share is generous: it holds at 79 percent only against a seven-round base, and Crunchbase prices ten rounds, which pulls the top two to about 74 percent. The $4.5bn addition is solid either way. The valuation multiples are honest arithmetic on one database's marks rather than anything tested by a transaction a third party can see.
The scorekeeper sells the scoreboard
The piece opens by pointing readers at Crunchbase's own Megadeals Board and then sources its valuations from Crunchbase, so the outlet keeping score also sells the score. Underneath that, each entry is an announcement whose timing and framing were chosen by a company and its incoming backers, and a tripled mark is exactly the figure both sides want on the record. Repeat funds compound it: Atreides and Valor each appear on more than one row of the same table.
Numbers firm, denominator soft
The individual round facts are precise, consistently reported and unlikely to be wrong. The weakness is one level up: a single publisher supplies all of them, and the share figures we built on top used a narrower base than the source's own list supports.