#AI chips

Google's Ironwood TPU has been generally available since April, yet there's still no public per-hour price. What that silence reveals about AI inference costs.

Google shipped Gemini 3.6 Flash, 3.5 Flash-Lite, and a restricted 3.5 Flash Cyber model. Why the gated cybersecurity variant is the real signal in the release.

Amazon's custom silicon runs at a $20B pace with $225B in Trainium commitments. What sold-out chips through 2027 mean for anyone buying AI training capacity.

CoreWeave measured 10× more DeepSeek-R1 tokens per megawatt on Vera Rubin NVL72, but missing methodology limits what NVDA investors can conclude today.

NVIDIA says its 32GB Jetson T3000 can match T5000 inference. TECHi examines the missing methodology, memory economics and 2027 hardware timeline for buyers.

Intel’s select-layer High-NA EUV use on Panther Lake creates a production-learning bridge for AI chips, without proving 18A economics are solved at scale.

A field guide to the four companies selling the machines behind AI memory — ASML, Applied Materials, Lam Research and KLA — and how the capex cycle pays them.

Lumentum jumped 12% to $794.30 on Meta's 14-gigawatt compute memo and is up ~115% in 2026. Inside the optics maker turning AI's bandwidth crunch into records.

Lam Research closed July 9 at $353.17, up 6%, as four banks raised targets to $400–$475. Up 154% in H1, LRCX sells the tools every memory fab must buy.

Qualcomm is buying Chris Lattner's Modular for $3.9B to attack NVIDIA's CUDA moat with a hardware-agnostic AI layer. Why it's the right target, and so hard to crack.

Micron reports fiscal Q3 on June 24. Why Wall Street doubled its price targets on a triple-digit DRAM surge — and what the print must deliver to justify it.

Entegris and KLA jumped as the semiconductor-equipment complex ripped. Why the deepest layer of the AI trade is booming — and the catch after the run.

Cerebras reports its first results as a public company on June 23. The per-share loss matters less than its $24.6B backlog and ~86% UAE revenue mix.

Nvidia and SK hynix signed a multiyear HBM4 memory pact for Vera Rubin. Why memory is AI's real bottleneck — and what it means for NVDA and Micron (MU).

Google reportedly ordered 3 million TPUs from Intel for 2028 and Nvidia is testing 18A. What the foundry wins mean for INTC stock — and the valuation gap.

Onto Innovation is an AI chip infrastructure stock tied to metrology, inspection, HBM packaging, and gate-all-around process control.

Oracle wants to own AI cloud capacity. Broadcom wants to tax custom silicon. Here is the ORCL vs AVGO setup before June earnings.

AMD stock reopens after Memorial Day with a new test: whether Venice, Helios and AI racks can justify the post-earnings re-rating.

Broadcom stock faces a Tuesday open test after the holiday. AVGO must prove it still beats Nvidia, Marvell and other AI alternatives.

Nvidia reports Q1 FY2027 today. The real test is not only revenue, but AI demand, bottlenecks, financing and customer ROI.

Broadcom stock faces a new 2026 test: whether AVGO's gigawatt AI deals deserve software-like valuation or a hardware-cycle discount.

Cerebras priced CBRS at $185, opened at $350, then cooled. Here is the real AI-chip bull case, the risks, and what investors should watch next.

Intel's 18A ramp is now an economic test: Foundry revenue is rising, losses remain heavy, and external customers still need proof.

Nvidia fell on 4 of its last 5 earnings beats. The May 20 Q1 FY27 print won't move the stock — only a Q2 guide above $86B will. Here's the math.