Daily Alpha · X
· Post-Market Alpha · by Buzzberg Research
X leaned constructive on memory and power but split on rates; the highest-value items were Micron's contract structure, insider selling at Cerebras, the Synopsys-OpenAI agreement, a Treasury supply debate, and Google's aggressive Argon pricing.
Themes on this desk
Micron after the print
Strategic customer agreements now cover over 35% of revenue through 2030 with $32B of mostly cash commitments and roughly $150B of remaining performance obligations; bulls flag that a 14-week FQ4 versus a 13-week FQ1 makes the guide look fl
Rates and term premium
The 10-year hit 5.304% and the 30-year 5.64%; the PCE methodology change may mechanically cut core inflation by up to 20bp; refinancing math implies roughly $30T of issuance through 2030, while Constan argues coupon issuance has been roughl
AI power and equipment
FERC accepted PJM's plan to procure 6.8GW of new capacity but suspended it to February 2027; BlackPantherCap argues that favors already-secured power names including IREN, NBIS, CIFR, WULF and BE. Constellation and Amazon signed a 20-year n
Ticker heat
?
Top voices by smart followers and alpha score
Market Radar →Jukan questions whether Synopsys-Amazon deal shifts custom silicon control to Annapurna
Jukan asks whether Amazon's Annapurna Labs will have more control following Synopsys' $1B+ multi-year custom silicon deal with Amazon (covering Nitro, Graviton, Trainium, with a license-plus-royalty model), and whether that would be bad news for Marvell and Alchip.
Does this mean Annapurna Labs will have more control? Would that be bad news for Marvell and Alchip?
If hyperscaler custom-silicon programs are being pulled in-house or re-intermediated, third-party ASIC design partners like Marvell and Alchip face share-loss risk even as total AI silicon spend grows.
Watch Whether Marvell or Alchip disclose changes to Amazon-related design win pipelines, and whether Synopsys' royalty model expands across other hyperscalers.
Source →Coupon issuance flat for ~3 years, challenging supply narratives
@dampedspring argues coupon issuance (roughly 7-year WAM) has been exactly the same for the last 2.5–3 years, pushing back on claims of a coupon supply surge.
Coupon issuance (roughly 7 year wam) has been exactly the same for the last 2.5 years. Call it 3.
If coupon supply has been stable, narratives blaming long-end yield pressure on a coupon deluge are misdiagnosed, changing the read on duration risk and Treasury market dynamics.
Watch TBAC issuance data confirming flat coupon issuance over the trailing 2.5–3 years versus claims of a surge.
Source →Synopsys-OpenAI partnership undermines the frontier-AI disruption short thesis
Firstadopter points to a Synopsys-OpenAI strategic partnership, a multi-year agreement to jointly develop GPT-Synopsys optimized for Synopsys EDA tools with revenue sharing and go-to-market collaboration, arguing that anyone short SNPS on frontier-AI disruption no longer has a valid thesis.
SNPS) and OpenAI today announced a strategic partnership that brings together OpenAI's advanced AI capabilities with Synopsys' industry-leading EDA tools
Directly challenges the bear case that AI models commoditize EDA, a potential re-rating catalyst for SNPS.
Watch Whether SNPS shares re-rate on the announcement and whether the GPT-Synopsys product reaches commercial availability.
Source →Cerebras COO sold $78M, leaving zero stake, after OpenAI reportedly skipped its chips
A quoted bubble boi post says Cerebras COO Dhiraj Mallick sold $78 million of stock, leaving him with zero stake, shortly after OpenAI's devday announced GPT6 ultrafast which SemiAnalysis reported does not use Cerebras wafer-scale engines; the poster reads it as lost confidence and more trouble ahead.
> bubble boi (@bubbleboi) > Last night Dhiraj Mallick, the chief operating officer of Cerebras sold $78 million dollars worth
A full exit by the COO combined with a reported loss of a flagship customer is a concrete negative signal for Cerebras' demand narrative and any pending valuation.
Watch Cerebras filings confirming the sale and any customer or partnership disclosure that offsets the reported OpenAI loss.
Source →Google launches Gemini 4 Argon at $2/$10 per million tokens, undercutting rivals
financialjuice reports Google set Argon pricing at $2 per million input tokens and $10 per million output tokens, following Pichai's introduction of Gemini 4 Argon.
Google sets Argon pricing at $2/M input, $10/M output tokens. $GOOGL > FinancialJuice (@financialjuice) > Alphabet's Pichai:
Aggressive token pricing on a frontier model pressures rival AI labs' margins and could accelerate enterprise adoption of Google's stack, reinforcing the price-war dynamic in frontier AI inference.
Watch Whether competitors match or undercut the $2/$10 pricing and whether Google discloses adoption or API volume metrics for Argon.
Source →