Monday, November 2, 2026

Marvell has agreed to buy Lumentum, the laser and optics maker, at an implied value of $115 billion.

It's all stock: each Lumentum share becomes 4.5 Marvell shares. With no cash involved, CNBC notes, the final value depends on Marvell's share price.

Why buy? Lumentum makes lasers and optical transceivers. Together, the companies say, they can lead in AI connectivity, and they project over $1 billion a year in cost savings. That is their own projection.

Investors were cautious. Today, CNBC reports, Marvell's shares fell 6%, and Lumentum's fell 11%. After the deal, Marvell's holders would own about 67% of the combined company, and Lumentum's about a third.

Closing is expected in the first half of 2028. It still needs approval from both companies' stockholders, and from regulators.

So, for you: the deal is more than a year from closing. Watch the stockholder votes and the regulators' reviews.

DailyChat is AI-generated, with an AI voice. Not investment or professional advice. See you tomorrow.

Anthropic's updated IPO filing: about $17 billion in third-quarter revenue, and a $23.3 billion net loss.

Anthropic filed the amended S-1 today, adding the third quarter, CNBC reports. The first version covered only the first half of 2026.

Revenue is up about 1,127% from a year earlier. But growth from one quarter to the next slowed to about 47%, from 144% in the second quarter.

For the first 9 months, the net loss was $104 billion. CNBC says most of that came from changes in the fair value of convertible notes and derivative liabilities.

And the commitments: purchase and lease agreements of about $895.4 billion, substantially all for compute and infrastructure, the filing says. CNBC's report doesn't say over how many years.

So, for you: Anthropic is preparing its road show, per CNBC. Its report gives no IPO date or price.

DailyChat is AI-generated, with an AI voice. Not investment or professional advice. See you tomorrow.

Meituan's LongCat-2.5 is out via API: a 1.6-trillion-parameter model for long agent tasks.

It's a mixture of experts, with about 48 billion parameters active per token. Meituan says it sees images natively, for visual coding and 3D tasks. Its site says it works with Claude Code and Codex.

The scores are from Meituan's own table. On BrowseComp, a browsing benchmark, it reports 91.9, against 90.8 for Claude Opus 5. On Terminal Bench 4.0, it trails: 38.5 against 53.9.

One question we checked: does it hold up on long runs with hundreds of tool calls? The blog says it was trained on long tasks, and on tasks built from real agent failures. It doesn't report how success changes on long or interrupted runs.

So, for you: try it through the API. The blog marks the open weights as coming soon and gives no date.

DailyChat is AI-generated, with an AI voice. Not professional advice. See you tomorrow.