Micron's $10B U.S. Bet, Anthropic's Chip Pivot, Memory at 50%

Micron is doubling down on domestic manufacturing, Anthropic is looking to build its own silicon supply chain, and memory chips have quietly become the backbone of the entire semiconductor industry. Meanwhile, rate hike expectations rattled tech stocks on Friday.

Key points

• Micron announced an additional $10 billion investment in U.S. semiconductor research and manufacturing, expanding its already substantial domestic footprint.
• Memory now accounts for 50% of global semiconductor revenue, a historic milestone driven by AI infrastructure demand, though analysts warn the pattern echoes a painful 2018 glut cycle.
• Nvidia CFO Colette Kress disclosed $279 billion in supply commitments for memory chips during Q2 earnings, signaling how aggressively the company is locking in capacity for its data center buildout.
• Anthropic is in talks with chip startup MatX on a partnership to accelerate custom chip design, pulling back from earlier acquisition discussions.
• OpenAI notified SpaceX it plans to end its Cursor agreement following SpaceX’s acquisition of the AI coding tool, citing concerns about how the technology will be used.

Why Micron’s bet matters now

Micron’s additional $10 billion U.S. commitment lands at a moment when memory has become anything but a commodity afterthought. Half of all global semiconductor revenue now flows through memory, a share that would have seemed implausible just a few years ago. AI training and inference workloads are the driving force, demanding high-bandwidth memory at a pace that has reshaped the entire supply chain.

The catch, flagged by analysts, is that this trajectory has a historical precedent that did not end well. The 2018 memory cycle saw a surge in capacity investment followed by a brutal oversupply collapse. The question now is whether AI demand is structural enough to absorb the wave of new capacity coming online, or whether producers are once again building ahead of a cliff. Micron’s move suggests the company is betting firmly on the former.

Nvidia’s $279 billion supply commitment provides some reassurance on the demand side. When the world’s dominant AI chip company locks in that much memory capacity, it signals that at least one major buyer is planning for sustained, large-scale consumption. Investors in memory-exposed names like Micron will be watching closely to see whether that demand materializes as steadily as the commitments imply.

Anthropic’s chip strategy shifts from acquisition to partnership

Anthropic’s talks with MatX represent a meaningful shift in how the Claude developer is approaching its hardware ambitions. Earlier discussions apparently centered on acquiring the startup outright. The current direction favors a partnership structure, which would let Anthropic accelerate custom silicon development without absorbing the full cost and complexity of owning a chip company.

The strategic motivation is straightforward: Anthropic’s Claude model family is scaling rapidly, and heavy reliance on Nvidia’s supply chain creates both cost and availability risks. Custom chips tailored to specific inference workloads can meaningfully reduce the cost per token at scale. Google, Meta, Amazon, and Microsoft have all pursued this path in various forms. Anthropic joining the list, even through a partnership rather than full vertical integration, fits the broader industry trend of AI labs diversifying away from a single chip supplier.

For Nvidia investors, each new custom silicon effort from a major AI lab is worth monitoring. None have displaced Nvidia from the training market, but inference at scale is a different equation, and that is where cost pressure is most acute.

Rate hike fears add macro pressure to AI names

Friday’s macro backdrop complicated the picture for tech broadly. Fed Chair Kevin Warsh’s remarks at Jackson Hole pushed traders to reprice rate expectations sharply higher. The probability of a September hike jumped from roughly 35% to nearly 58% by end of day, per CME Group data. The two-year Treasury yield rose to 4.35% from 4.22% in a matter of hours.

The Nasdaq slipped 0.5% on the session, with the S&P 500 down 0.2%. Marvell Technology fell 10.3% despite reporting earnings that beat estimates, a reminder that elevated valuations across the AI sector leave little room for disappointment. Stocks that have already priced in strong AI revenue growth are particularly exposed when rate expectations shift, since higher discount rates compress the present value of future earnings. For investors in AI names trading at stretched multiples, the macro environment heading into September deserves attention alongside the company-level fundamentals.

One side note on influence lists

Jensen Huang’s absence from Time’s 100 most influential AI figures for 2026 generated attention on Friday. Whether the omission reflects editorial judgment, shifting narratives around Nvidia’s dominance, or something else is speculative. It does not change Nvidia’s competitive position or financial outlook. But it is a small data point in the ongoing conversation about whether the AI chip supercycle is entering a new phase, one where the picks-and-shovels narrative gets harder to sell without scrutiny.

Nothing in this article constitutes investment advice. This site is independent and not affiliated with any company or broker mentioned.