On October 26 Itaú Unibanco disclosed that it will buy NeoSpace for R$2.925 billion, roughly US$540 million at that day's exchange rate, with part of the price deferred over five years and subject to conditions. The bank was already in the room: it led NeoSpace's US$18 million round in January 2025, has used NeoSpace's models on its own data for nearly two years, and bought the founders' previous company in 2020. Capital Aberto reported that when Itaú signed, NeoSpace was negotiating a new round, not yet closed, at a US$1.25 billion valuation.
The bank bid before that outside price was tested, and the rest of the week showed the other side of the same mechanism. Datadog, which Globes reported in 2025 was in talks to buy Upwind, offered no premium to the $3.8 billion its venture investors set, TechCrunch reported, though Calcalist reported no offer was made; Upwind raised $400 million at that price and stayed independent. Salesforce agreed to buy Rune with no reported prior tie, and PitchBook finds much of the IPO backlog staying private on private money, even as Anthropic filed to list. A buyer already in the room can bid before anyone outside has priced you. A founder's leverage is set when that investor or customer first comes in, and in a round they could raise without it.
Itaú Unibanco disclosed on October 26 that it will buy all of NeoSpace for R$2.925 billion, roughly US$540 million at that day's exchange rate, with one installment at closing and the rest deferred over five years, subject to conditions and to approval by Brazil's antitrust regulator and central bank. The bank was in the room three ways: it led NeoSpace's US$18 million round in January 2025, taking about 15%; NeoSpace has spent nearly two years building models on Itaú's own data for credit, CRM and fraud; and Itaú bought the founders' previous company, Zup, in 2020. Capital Aberto reported that when the bank signed, NeoSpace was negotiating a round of up to US$250 million at a US$1.25 billion valuation that had not closed. The headline says the investor became the exit. The real signal is timing: the buyer that knew NeoSpace best signed before any outside investor's price was tested.
Upwind, which secures cloud and AI workloads while they run, raised a $400 million Series C at a $3.8 billion valuation on October 28, led by Bessemer Venture Partners, up from about $1.5 billion in January. Globes reported in July 2025 that Datadog was in advanced talks to buy Upwind. TechCrunch, citing a person familiar with the company, reported that Datadog tried again with no premium above the $3.8 billion its venture investors set, and that Upwind chose to stay independent; Calcalist had reported on October 22 that the offer was never made. If the TechCrunch report is right, private money at the same price is what let the company say no.
A Crunchbase News guest column on October 29 by MGV's Marc Schröder, citing Crunchbase data, counted 272 US Series A rounds in the third quarter against 377 a year earlier, and found that only 16% of the 2024 cohort of seed companies that raised $1 million or more has raised past seed, against 55% or more through 2020. Uncork's Andy McLoughlin is quoted saying Series A investors now expect $2 million to $3 million of ARR. The 2024 cohort has had less time than earlier ones, so part of that gap is age; the direction still matters. If about one in six recent seed companies has raised again, investors who would fund you at an insider's price may be scarcer, and an insider's offer may have less to beat.
Salesforce agreed on October 29 to acquire Rune Technologies, the startup founded in 2024 by Anduril alumni, whose TyrOS logistics software is deployed with the Army and Marine Corps; no prior stake or partnership between them has been reported. TechCrunch, citing Resilience Media, reported a price of $375 million in cash and stock, against a $127 million valuation at Rune's Series A, per PitchBook; Salesforce did not disclose terms.
DoorDash signed a definitive agreement on October 28, announced the next day, to acquire Instacart, a listed company, for $65 a share, about $17 billion in equity value and a 26% premium to its October 22 close, at least 75% in cash. In 2021 the two held merger talks, initiated by Instacart, at a likely $40 billion to $50 billion, The Information reported then; those talks collapsed partly over antitrust concerns. The buyer that was in the room in 2021 came back at well under half that range, but 26% above Instacart's share price: the gap is the market's, not the insider's. Closing is expected between the end of 2027 and the first half of 2028.
Intel agreed on October 26 to acquire Classiq, an Israeli quantum-software company, in a deal reported at $2.2 billion to $2.3 billion; Intel did not disclose terms. Calcalist reported that Intel's chief executive, Lip-Bu Tan, had invested personally in Classiq's 2022 Series B and recused himself from the talks, while Classiq's corporate investors included Intel's rivals AMD and Qualcomm.
PitchBook reported on October 30 that of 254 companies its exit model expected to go public by the end of 2025, 53 did, 89 have faded as candidates and 112 are still waiting, and just 58 still look likely to list within three years. Its read is that much of the backlog is "holding out, not locked out": staying private because private investors keep funding it. Private money, not a listing, is the second path much of this backlog is using.
Anthropic's public registration statement, filed October 26 for a Nasdaq listing under the symbol ANT, says model versions released in the six months to September 30 earned over 80% of its third-quarter revenue. It reports net revenue retention of about 900%, measured as one quarter against the same quarter a year earlier and excluding Claude Pro and Max subscribers. The second quarter was the first in the filing with a GAAP operating profit, $974 million, after an $8.06 billion operating loss in 2025; the filing does not report a third-quarter operating result. For a startup built on it, the model vendor is a supplier whose revenue depends on replacing the version you built on; plan for that as you would for any supplier with that much power.
Reuters reported on October 28, citing two people, that the Justice Department is asking whether Nvidia's agreement to buy Hugging Face, signed September 2 for about $11.9 billion plus up to $1 billion in retention equity, would let it "capture a fundamental distribution point for AI". Nvidia was already an investor, one of eight corporate backers in Hugging Face's 2023 Series D. No formal second request has been disclosed, and Nvidia committed in its September 8-K to keep the platform open to other chipmakers.
The FT reported on October 30, the day of OpenAI's investor commitment deadline, that it had cut its new round to $20 billion from $30 billion; the Wall Street Journal reported about $11 billion secured so far, and the FT said a later offering to wealthy individuals could bring the round to its new target. OpenAI has not commented, and nothing has closed. Private money is the second path for the largest companies too, and this week even OpenAI had to work for it.
Alphabet reported third-quarter free cash flow of −$9.1 billion on October 28, its second negative quarter in a row, as $51.4 billion of capital spending outran $42.3 billion of operating cash flow. Meta priced $25 billion of stock under forward sale agreements and about $12.5 billion of mandatory convertible preferred on October 26, and S&P moved Meta's outlook to negative on October 30. Two of the companies best placed to buy startups are putting their cash, and new money, into data centers. That does not prove they will buy fewer startups, but if compute comes first, fewer outside bidders may be ready to compete with the insiders already in your room.
The Information's AI Agenda column argued on October 26 that rounds split into tranches at different valuations, once unusual, are now standard in AI, and, citing unnamed investors, that corporates such as Nvidia often take more than a quarter of the pricier tranche. Bloomberg reported on October 30 that Baseten is in talks to raise about $2 billion in two tranches at $26 billion and $30 billion, with Sands Capital expected to lead; nothing has closed. A corporate paying the higher tranche is not a sign it will pay more to buy you; it may be what it pays to get into the room. A headline built this way is a price some investors paid, not the price an acquirer has to beat.
The Prime Minister's Office said on October 29 that Canada will make a staged equity investment of up to C$1.6 billion in Cohere, tied to commitments of at least 1,000 new technical jobs, at least C$1.7 billion of Canadian compute by 2030, and keeping its headquarters and incorporation in Canada; Nvidia, Bell and Schwarz Group co-invest. BetaKit reports it is part of a Series E targeting up to C$7 billion at a valuation of nearly C$36 billion, which had an initial close the same day. The compute commitment alone is larger than Canada's check; an investor that sets where you must stay and spend also shapes who can buy you.
Itaú buying NeoSpace reads like the cleanest exit there is: the investor and customer that knows you best decides to own you. I read it as a warning. The buyer who reads your monthly numbers knows how long you can afford to say no.
My #034 ↗ issue argued for choosing a first check that still makes sense at the second. NeoSpace was negotiating an outside round, Capital Aberto reported, when its lead investor signed to buy it for roughly US$540 million, part of it deferred and conditional. That round had not closed: it was a second path NeoSpace had not yet secured. Upwind had secured one; whatever Datadog did or did not offer, and the reports disagree, $400 million at $3.8 billion let it stay independent.
The leverage is set when a strategic investor or large customer first comes in, not when it bids. At that term sheet, limit its information rights to what your other investors get, keep it out of board talk of a sale, and refuse it a right to match offers for the company. With no second path, ask your investors in writing whether they would fund you at the offer price, then negotiate how much is paid at closing. Deferred money tied to who stays moves value from your other holders, including option holders who have left, to whoever remains.
Build the second path the year the strategic invests, not the week it bids. Belief becomes capital when you have built something an outsider would pay for.