💵AI startups use split valuations to hit $1B headlines

Serval was valued at $235M in August 2025. In December it announced a $75M round at a $1B valuation led by Sequoia. According to WSJ, Sequoia actually invested at a $400M valuation.

The tactic: offer the lead investor a lower valuation, then close additional capital days later at a higher one. In Serval’s case, Sequoia came in at $400M, while the headline round priced the company at $1B. On paper, the lead’s stake immediately marked up.

WSJ cites another case, Aaru, where part of the round was done at $450M and part at $1B. Carta data shows around 20 similar deals in the past year, with activity rising in late 2025.

As the AI funding market cools, valuation engineering is becoming part of the playbook.

✔️ Subscribe on @venture

🚨 Uber CFO buys 22,400 shares on the open market

Uber CFO Prashanth Mahendra-Rajah purchased 22,400 shares at about $71.25 per share. The filing was made on Feb 24, 2026. The transaction was an open market buy, not options.

The stock is trading roughly 30% below its recent highs.

When a CFO deploys personal capital at these levels, it stands out.

✔️ Subscribe on @venture

🚨 Block cuts 4,000+ jobs as Jack Dorsey pivots to AI-first model

Jack Dorsey announced Block will reduce headcount from over 10,000 to under 6,000 employees. More than 4,000 people are leaving or entering consultation.

Dorsey said the move is not due to weak business performance. Instead, he pointed to AI tools and smaller, flatter teams enabling a new way of building and running the company. He chose a single large cut over gradual reductions.

Severance includes 20 weeks of pay plus one week per year of tenure, six months of healthcare, vested equity through May, and $5,000 in transition support.

This is one of the first large scale workforce reductions explicitly tied to an AI-driven operating shift.

✔️ Subscribe on @venture

Just In: OpenAI raises $110 billion at a $730 billion valuation.

✔️ Subscribe on @venture

🧠 Jensen Huang rejects “SaaSpocalypse” as AI agents layer on top of software

Nvidia CEO Jensen Huang said the idea of a “SaaSpocalypse” is wrong. In his view, AI agents will not replace software tools but operate on top of them.

He argued that value will accrue to the platforms and workflows agents integrate with, rather than triggering a full wipeout of SaaS vendors.

The shift may be additive, not destructive, for existing software stacks.

✔️ Subscribe on @venture

Warren Buffet's biggest flex about being rich.

✔️ Subscribe on @venture

✔️ Subscribe on @venture

🤖🤖 Pentagon drops Anthropic, signs new deal with OpenAI

Anthropic refused to fully accept U.S. Department of Defense terms and stated Claude would not be used for autonomous weapons or mass surveillance. The Pentagon then terminated a reported $200B contract and labeled the company a supply chain risk.

Federal agencies are required to phase out Claude within 6 months. President Trump publicly criticized Anthropic and called for agencies to stop using its technology.

At the same time, OpenAI signed a new agreement with the Department of Defense to deploy its models inside the Pentagon’s secure cloud network for national security use. OpenAI will form a Forward Deployed Engineer team to support on site deployment.

Both companies reportedly insisted on restrictions around mass surveillance and human oversight in use of force decisions. Only one of them kept the contract.

✔️ Subscribe on @venture

🔥 OpenAI closes $110B private round at $730B valuation

OpenAI has completed a $110B private funding round at a $730B pre money valuation, implying about $840B post money. Part of the funding will come in services rather than pure cash.

SoftBank and Nvidia committed $30B each. Amazon committed $50B plus a strategic partnership. Of Amazon’s amount, $15B is available upfront, while $35B is tied to conditions such as OpenAI using a required volume of Trainium chips.

The new Nvidia investment appears to replace the previously discussed $100B agreement between the companies. SoftBank financed part of its commitment by selling $5.8B in Nvidia shares.

✔️ Subscribe on @venture

⚔️ US used Claude in Iran strikes hours after Trump called for a ban

The Wall Street Journal reports that US Central Command used Claude during recent strikes on Iran.

The system supported intelligence assessments, target identification, and combat scenario modeling.

Formally, this does not violate Trump’s order. Agencies have 6 months to phase out Claude.

The timing stands out.

On Friday, Anthropic refused to accept Pentagon terms. The same day, Trump urged agencies to stop using Claude and called Anthropic “radical.”

Hours later, Claude was used in a live military operation.

✔️ Subscribe on @venture

🎤 Claude hits #1 in US App Store after Pentagon clash

Amid the conflict between Anthropic and the Pentagon, Claude climbed to #1 in the US App Store.

Social feeds filled with screenshots of users canceling ChatGPT and switching payments to Anthropic.

The surge followed public tension around defense contracts and political criticism.

The episode pushed Claude into mainstream attention at scale.

✔️ Subscribe on @venture

📊 YC’s Garry Tan: half of AI agent usage is in software, most verticals are open

Anthropic’s recent research on AI agents shows that 49.7% of agent tool calls are in software engineering. The next category, back office automation, is at 9.1%. Most other domains sit below 5%.

Healthcare is at 1.0%, legal at 0.9%, finance and accounting at 4.0%. Many industry verticals barely register in current agent deployment.

YC’s Garry Tan argues this leaves large parts of the market underpenetrated. While software agents are crowded, sector specific applications remain wide open.

The data suggests the next wave may be vertical, not horizontal.

✔️ Subscribe on @venture

🎰 OpenAI fires employee over insider trading on Polymarket

OpenAI dismissed an employee for insider trading tied to bets on Polymarket. It appears to be the first confirmed big tech firing linked to prediction markets.

Before the launch of the ChatGPT browser, 13 new Polymarket accounts appeared 40 hours ahead of release.

Together they placed $300k on the correct outcome and reportedly made a strong return.

A similar case happened in 2023 during Sam Altman’s removal, when a new trader bet on his comeback and earned $16k before disappearing.

Insider concerns around product launch dates had been building. Prediction markets are starting to test corporate information control.

✔️ Subscribe on @venture

🚀 SpaceX plans 150 Mbps satellite internet for smartphones

SpaceX already offers satellite connectivity for smartphones through T-Mobile, with current speeds around 4 Mbps. That is enough for text messaging and low-quality calls.

By the end of 2027, Starlink is expected to upgrade the network to peak speeds of up to 150 Mbps. Real-world speeds will be lower, but high enough for video calls and content streaming.

✔️ Subscribe on @venture

📈 AI startups are growing fast. Until they aren’t.

A lot of teams are building guardrails, adapters, and workflows on top of today’s models. It works. Revenue grows. Real users.

The bet is that base models will always need that extra layer. But if future versions handle hallucinations, memory, and routing on their own, that middle layer shrinks fast.

The turkey looks fine for 1,000 days. The risk sits on day 1,001.

✔️ Subscribe on @venture