Institutional-grade equity research on AI infrastructure, power, and the capital cycle. Synthesized from the same data pipeline that feeds the MCP endpoint.
The pass-through gap is an accounting problem, not a demand collapse. At 3.3x forward earnings, the market is pricing in a recession the physical data doesn't show.
Five machines in 2026 against ASML's 130. Five generations behind. Zero precedent of Chinese lithography running production. The supply story hasn't changed.
Positioning data says the unwind is fading. SK Hynix reports tomorrow and the Street still forecasts a record ₩64 trillion operating profit.
The Q3 guide missed by a mile, but the weakness is a factory move, not demand destruction. Computing is surging, margins are expanding, and the sell-off from $96 to $46 looks like an overreaction to the wrong story.
MSFT reports FQ4 Tuesday with $37B of AI revenue already booked, Azure growing 40%, and the Street calling for 46% upside. The stock needs one thing: proof the capex is working.
The Liang Wenfeng transcript reveals a 10-month GPU payback at 85% margins — and open-source just captured three-quarters of all AI token volume
The Bargo disruption model runs 32 software companies through three guardrail tests. The AI-vs-fundamentals correlation collapses to zero under basic controls. The seat-subscription model is under genuine pressure — but AI exposure doesn't predict which names struggle.
The per-token price debate misses the point. K3 is a 2.8T-parameter brute with an immature inference stack that burns three times the tokens of its peers. The open-weight release today is a compute-demand story.
Both are right about their champion. Neither addressed that China is winning.
Near Fable 5 intelligence at half the price — and beating it on Frontier-Bench, ARC-AGI-3, and OSWorld. The product ladder is complete.
The "open source with impunity" thesis sounds clever. The hardware reality says something else entirely.
A coalition of 25 companies led by NVIDIA and Microsoft is asking Washington not to kill open-weight AI. OpenAI and Anthropic didn't sign. Here's what's really at stake.
A deterministic supply-and-demand model of AI inference, anchored to what the labs actually disclose
Half the sticker price, twice the tokens burned — how Kimi K3's apparent cost advantage evaporates under scrutiny
The Big Short investor halved his bearish bet in late June before reposting his takedown. The options market is pricing a battleground, not a collapse.
The 0.99 correlation that turns every NVIDIA earnings print into a 9-month forecast for AI infrastructure spending — and what tonight's GOOGL print means for the buildout trade
The gap between American and Chinese AI models is real and widening. But the story of who wins and loses is more complicated than a 60x price difference suggests.
IBM just proved the budget-crowding-out thesis. But the market already priced it in. Now ServiceNow and IBM report Wednesday, and the whole SaaS narrative hangs in the balance.
The viral claim that K3 runs on 277KB of SRAM is marketing. The real story: frontier MoE models need more memory than ever, and every layer of the hierarchy is supply-constrained.
Washington finally cleared H200 sales into China in July 2026. Beijing had told its firms to stop buying six months earlier.
Chip stocks fell into a bear market this month. The companies building the physical plant behind AI kept beating their own guidance.
H200 and B200 spot prices are rising, token demand just spiked 38% in a week, and hyperscalers are spending $725 billion this year. But the overall market is still balanced, and H100 remains loose.
Goldman says 250,000 humanoids ship in 2030. Bank of America says 1.2 million. That 5x gap between serious desks is the whole argument for owning what goes inside the robot instead.
China made 94% of the world's sintered permanent magnets in 2024. Every humanoid needs 3.5 kg of them. Musk already had Optimus throttled by the export rule, and the US answer just got blacklisted by Beijing.
Token consumption hit 58.5 trillion per week after GPT-5.6 and DeepSeek V4 launched in July. H200 rental capacity is already tight. The Jevons effect is real.