💰 SoftBank Pledges OpenAI Shares for Loans

SoftBank owns about 13% of OpenAI and has invested roughly $64.6 billion, mostly not its own cash. It sold $6 billion in Nvidia shares to fund the initial stake without debt.

In March, SoftBank took a $40 billion bridge loan at 8%, paying it to OpenAI but due in a year. In April, it borrowed another $6 billion using its OpenAI shares as collateral.

If OpenAI’s IPO delays or valuation drops, SoftBank risks default and heavy losses. The company is uneasy, but Masayoshi Son ignores warnings.

SoftBank faced a similar fate in 2019 with WeWork, losing $10 billion after a failed IPO. This time, the stakes are 15 times higher.

✔️ @venture

🖥 Microsoft to drop Claude Code for Copilot CLI

In December, Microsoft made a deal with Anthropic and encouraged engineers to use Claude Code. The tool quickly gained popularity among developers.

Now Microsoft plans to cancel its Claude Code license and switch to Copilot CLI. Developers on major products like Windows and Microsoft 365 must move all data from Claude by the end of June.

✔️ @venture

🤖 Anthropic now outpaces OpenAI revenue

The Information estimates Anthropic currently generates at least 35% more annual revenue than OpenAI. Earlier this year, OpenAI led by nearly $1 billion.

Anthropic develops AI models similar to OpenAI's, focusing on large language models and AI safety.

Growth at Anthropic is accelerating while OpenAI sees slowing user growth. This shift could impact their upcoming IPOs.

✔️ @venture

🤖 Anthropic released Opus 4.8

The new version improves coding, agent workflows, and reasoning performance while keeping the same pricing: $5 per million input tokens and $25 per million output tokens.

Claude Code now supports dynamic workflows. The model can break large tasks into parts, run parallel subagents, and review results before responding.

Anthropic also says Fast mode for Opus 4.8 is 2.5x faster and 3x cheaper than previous models. The company claims the model is also about 4x less likely to miss bugs in its own code.

✔️ @venture

🚖 Uber questions AI spending amid rising costs

Uber's COO said heavy AI spending is hard to justify as token consumption grows without clear consumer benefits. The company used its annual AI budget in just 4 months.

Nvidia's VP noted their compute costs now exceed employee expenses, showing AI is becoming more expensive than human labor.

✔️ @venture

🤖 Anthropic nears $1T valuation after $65B raise

Anthropic closed a $65 billion Series H round, raising its valuation to $965 billion from $380 billion months ago.

It reported an annual recurring revenue of $47 billion, surpassing peers including OpenAI, valued at $852 billion.

✔️ @venture

🚫 China is restricting overseas travel for top AI engineers

China has started limiting foreign travel for AI researchers and engineers working at private companies like Alibaba and DeepSeek, according to Bloomberg.

People involved in advanced AI work now reportedly need government approval before leaving the country. The restrictions target specialists considered strategically important for China’s AI race with the US.

The move shows how seriously Beijing now treats AI talent and research. Engineers are increasingly being viewed less like employees and more like critical infrastructure.

The global AI race is starting to look closer to semiconductor policy than normal tech hiring.

✔️ @venture

$2,600 invested in Tesla’s IPO is worth more than $1 million today.

✔️ @venture

“When people start telling you you’re crazy, you just might be on to the most important innovation in your life.”

~ Larry Ellison

✔️ @venture

🔋 Google engineer allegedly made $1.2M using insider information on Polymarket

The US Department of Justice charged Google engineer Michele Spagnolo with making $1.2 million from prediction market bets using non-public company information.

According to prosecutors, Spagnolo used advance knowledge of Google’s annual Year in Search rankings to place bets on Polymarket before the results were published. He reportedly wagered more than $2.7 million in total.

Google said the information came from internal marketing materials that were widely accessible to employees, but using it for betting was prohibited.

✔️ @venture

⚖️ AI is pushing more people to sue without lawyers

MIT and USC researchers analyzed 4.5M federal cases and 46M PACER records to study how generative AI is changing US courts.

For almost 20 years, only ~11% of federal cases were filed without attorneys. In 2025, that jumped to 16.8%.

The increase is strongest in paperwork-heavy cases. AI helps people draft complaints, handle legal formatting, and navigate filings without paying a lawyer.

Courts are already seeing the effect. These cases are not resolved faster, but they generate far more activity inside the system. In the first 180 days, court filings and procedural records in self-represented cases rose 158% versus the pre-AI period.

Researchers also found AI-written complaints rising fast: from ~1% in 2023 to 18% by early 2026.

✔️ @venture