On November 5 Carta reported that companies on its platform raised $49.3 billion across 1,179 rounds in the third quarter, part of the strongest six months of fundraising since late 2021. Median valuations were higher than in 2025 at every stage, and median dilution kept falling. The number of rounds did not follow the money: total deal count across all stages was down 21% from a year earlier. Carta's own question for founders: “will a startup be able to raise capital at all?”
Three days earlier Crunchbase had found the same split in North America: dollars up 50% year over year to $92 billion, deal counts down 16% to the lowest quarter since 2023, with the drop concentrated at seed and early stage. It counted seven companies raising just over half of all US seed and venture dollars in the first ten months of the year. PitchBook found the same concentration one layer up, with LPs shifting to large, established managers, and Sequoia shut Arc and its scout program, two of its earliest doors. The price is better for the rounds that close. Fewer of them closed, so plan around whether your round closes, not its valuation.
Carta's State of Private Markets report for the third quarter, published November 5, counts $49.3 billion raised across 1,179 rounds by companies on its platform; over the last two quarters they raised more than $100 billion, the strongest six months since the fourth quarter of 2021 through the first of 2022. At every stage, median valuations are significantly higher in 2026 than in 2025, median dilution is still falling, and the time between priced rounds is shrinking at most stages. The caveat is the count: total deal count across all stages was down 21% year over year, and against the first quarter of 2023, quarterly cash is up 157% while the round count is 8% lower, so the average round has more than doubled. The headline says the market is friendlier to founders than at any point in several years. The real signal is that those terms describe the rounds that closed, and fewer did.
Crunchbase reported on November 2 that North American startups raised $92 billion in the third quarter, up 50% from a year earlier, while reported deal counts fell 16% year over year and 14% from the second quarter, the lowest quarterly total since 2023. The contraction is concentrated in seed and early-stage deals; late-stage counts rose. Two days later Crunchbase counted seven companies, among them OpenAI, Anthropic, Anduril and Databricks, that raised more than $260 billion, just over half of all US seed and venture investment in the first ten months of the year. The headline dollar totals describe a handful of companies, not the market most founders raise in.
PitchBook's third-quarter US VC Fundraising and Returns Report, published November 4, puts US venture fundraising at $108.5 billion for the year so far, pacing for the strongest year since 2022, and calls it a very concentrated allocation. LPs have shifted their allocations to large, established managers, the authors write, while emerging managers “are being left to wilt,” and liquidity remains sluggish. It is the week's concentration one layer up: the money that funds venture firms is pooling with large, established managers, as Crunchbase found the money those firms invest pooling in seven companies.
In a memo posted on November 4, “The Sequoia of Tomorrow,” the firm said it is closing its London office (“We simply don't see the need for a local presence”), sunsetting Arc, its batch program, and its Scouts program, ending the split between its Early and Growth teams, and simplifying the Sequoia Capital Fund. One of venture's best-known firms said it has raised about $10 billion in new funds and closed two of its earliest doors in the same memo.
Creandum, the European early-stage firm behind Lovable, Klarna and Black Forest Labs, said on November 2 it had closed Fund VIII at $750 million, its largest, after a €500 million (about $560 million) Fund VII. The fund backs founders at seed and Series A, with half of the capital going to new investments and the rest to follow-on rounds, and Creandum has opened a new San Francisco office that it says gives founders a single partner across Europe and the US. Half of the fund, about $375 million, goes to first checks at seed and Series A, in Europe and, increasingly, the US.
On November 3, voters decided 41 local ballot measures on data-centre development, and in 40 of them the side limiting data centres won, Ballotpedia counted, including five outright bans, twelve partial bans and three six-month moratoriums, all three in Maine. Texas's governor-elect campaigned on a moratorium on data-centre approvals and takes office January 19; candidates in Ohio and Georgia also campaigned on restricting data-centre development, and Texas already paused grid and environmental permitting for data centres under Governor Abbott. Texas holds about 27% of planned data-centre gigawatts, by a Wells Fargo estimate Fortune cited. These votes add a constraint money does not settle: a local vote on where compute gets built.
Oracle priced $10 billion of notes on November 2 at coupons from 7.0% to 8.65%, alongside about $10 billion of new stock, at yields about 0.35 percentage points above its existing debt, Bloomberg reported. Meta sold $15 billion of dollar notes on November 3 plus about €9 billion in a debut euro deal; Bloomberg reported Meta had aimed to sell $25 billion of bonds this quarter, and its 2056 tranche priced about 0.9 percentage points wider than a comparable Meta note sold in April. No deal failed: buyers took the paper at higher yields, and Bloomberg reported issuers are deliberately keeping offerings smaller. The largest borrowers in the buildout still raised the money; they paid more for it.
Shield, the defence autonomy company known as Shield AI, announced on November 5 a definitive agreement to buy radar maker Echodyne in a deal valued at $5.3 billion, $4.8 billion at closing plus up to $500 million tied to performance targets, pending regulatory approval. To fund it, Shield is raising $2.5 billion in three layers: a $1 billion Series H at a $28.8 billion post-money valuation, $1 billion of structured equity from Blackstone funds, and $500 million of debt arranged by JPMorgan. Crunchbase's weekly list recorded $2 billion of equity and $500 million of debt as “secured”; the release says Shield “is raising” it. In a week of headline totals, it is a reminder to read what a total contains.
Nooks, which sells AI sales assistants and revenue agents, secured a $500 million commitment from General Catalyst's Customer Value Fund in exchange for a share of the returns its capital helps generate instead of an equity stake, chief executive Dan Lee told Reuters, as Silicon Report relayed on November 4. The money goes to sales and marketing; the return share, repayment terms and pace of drawing it were not disclosed. Capital like this is not a priced round, so round counts do not capture it: some of the money moving to proven go-to-market engines arrives outside the priced round entirely.
On November 4 Meta acquired Conductor, software that runs several AI coding agents in parallel, for undisclosed terms; its founder said he looks forward to working with the leaders of Meta Superintelligence Labs. Its core features keep working until early next year, but its cloud service shuts down on November 20, after which user cloud data will be deleted, TechCrunch reported. The same day OpenAI agreed to acquire Pierre, maker of Git-based code storage and open-source code-review libraries, pending regulatory approval, and said Code Storage customers can keep using the product after closing. A day later Dropbox announced plans to buy Tigris Data, which builds distributed data storage for AI workloads, for $350 million. In a week when the round count was the story, two developer-tool teams went to labs instead, one deal done and one pending, on undisclosed terms.
Bloomberg reported on November 6 that OpenAI has closed an initial portion of roughly $11 billion, wired to the company that Friday, in a round that values it at $1.25 trillion, according to people familiar with the matter; the broader round remains open and could reach as much as $20 billion in the coming weeks. In late September Bloomberg had reported that OpenAI was looking to raise $30 billion in this round, and TechCrunch notes the current figures fall well short of that target. Even at the top of the market, OpenAI's round is arriving in parts, and is tracking short of its first target.
Renaissance Capital counted 8 IPOs raising $1.2 billion in October, plus one direct listing, against a ten-year October average of 22 IPOs and $5.1 billion; nearly two-thirds of the month's IPOs finished below their issue price. The only US debut in the first week of November was a $91 million best-efforts community-bank offering, and Renaissance's IPO index stood 6.8% down for the year on November 5, against an 8.2% gain for the S&P 500. Few listings mean few IPO exits, and PitchBook's fund report the same week said liquidity remains sluggish.
Humble Robotics, which builds autonomous electric haulers for regional freight, raised a $51 million Series A led by Obvious Ventures on November 5, bringing its total to $75 million less than a year after founding. Guava, a developer platform for enterprise voice agents that trains and serves its own speech and language models, raised a $12.5 million seed led by Acrew Capital. Phaze, which pools corporate workstations for employees and computer-use AI agents, raised an $8.75 million seed led by a16z, whose Martin Casado said the firm was “excited to support them again”; the founders previously built Parsec. Each came with a specific claim: Guava trains and serves its own models, Humble builds its own haulers, and Phaze's lead had backed its founders before. Three closes are not a recipe; rounds that stalled with the same claims do not make the news.
Carta's third quarter reads like the best market for founders in years: higher valuations at every stage, less dilution, shorter gaps between rounds at most stages. Every one of those medians describes a round that closed, and 21% fewer closed than a year earlier.
Crunchbase found North American deals at their lowest since 2023, the drop concentrated at seed and early stage, and seven companies raising just over half of US venture dollars. PitchBook found LPs favouring established managers. A count is not your odds, since nobody counts who tried, but the dollars are pooling.
A wager I lost: in #033 ↗ I bet Anthropic's public filing would miss what investors were told by more than 5%. Its amended registration statement ↗, filed November 2, reports third-quarter revenue of $16.96 billion against the roughly $16.8 billion attendees said investors heard, after the second quarter held. I concede: definitions mattered less than I argued.
My next one: Crunchbase's first January report will put combined North American seed and early-stage deals for the fourth quarter at least 10% below the same quarter of 2025. If not, I will record the miss here and withdraw my read that those rounds are still thinning. No such count, no bet.
Until then, plan around the close: budget your team's runway for a first pass that misses, list realistic leads by name, and check what each has announced at your stage. If the list runs short, extend runway. Then take the fair price that closes; if fewer comparable companies in your market closed, yours buys a longer lead.