Daily Alpha · X
· Premarket Alpha · by Buzzberg Research
X concentrated on the conflict between new AI legal risk and still-tight physical infrastructure, with specialist posts providing the clearest stock-level evidence.
Themes on this desk
AI law versus demand
Lina Khan’s existing-law argument creates near-term liability risk, while current-model deployment and HBM scarcity argue that demand has not yet slowed.
Component bottlenecks
Murata’s high-end MLCC plan and OCC’s fiber shortage independently point to demand exceeding selected component capacity.
Ticker heat
?
Top voices by smart followers and alpha score
Market Radar →Murata VP: AI MLCC demand growth extends at least through 2028
Murata's Vice President says MLCC content per AI server is rising, with demand shifting toward small size, high capacitance and high heat resistance, and that AI-related MLCC demand growth is expected to continue at least through 2028, broadening from data centers into edge AI and physical AI/robotics.
growth is expected to continue at least through 2028 - Over the medium to long term, demand centered on AI
Confirms a multi-year AI-driven component demand runway for the leading MLCC supplier, supporting the high-end mix shift narrative for passive component makers.
Watch Murata order/backlog commentary and MLCC utilization rates through FY2027; any guidance cut to AI-related component demand would invalidate.
Source →Lina Khan: existing law already reaches AI firms and their CEOs
Former FTC chair Lina Khan argued there is no AI exemption from laws already on the books: unvetted model releases can violate consumer protection law, flawed AI tools without rogue-agent controls can be 'unfair or deceptive' under the FTC Act, state AGs are exploring criminal liability for AI firms and CEOs, and concentrated cross-investments (e.g., Nvidia's Hugging Face acquisition blunting an OpenAI liability suit) undermine accountability.
For example, OpenAI could face liability given the Hugging Face incident, but Hugging Face being bought up by Nvidia means
Introduces a legal/regulatory tail risk to AI platform economics and to the cross-holding structure of the AI stack, independent of the demand debate that drove the tape.
Watch Actual FTC/state AG enforcement actions or investigations naming AI labs, and any antitrust scrutiny of AI cross-investments.
Source →OCC's binding constraint is optical fiber supply, not demand
Management said the primary impediment to ramping manufacturing is optical fiber shortages, and stated these industry challenges will not prevent strong top-line revenue growth for the remainder of fiscal 2026.
“The primary impediment to ramping up manufacturing at the current time is really optical fiber shortages.” While there are certain
A supply-constrained producer facing industry-wide fiber scarcity signals pricing power and demand strength, but also caps near-term revenue upside regardless of order flow.
Watch Whether OCC reports revenue growth consistent with its FY2026 guidance and whether fiber availability improves or worsens next quarter.
Source →IREN valuation challenged: $110B scenario needs $4.4B net income, leaving little for depreciation and interest
@jpinsights says he is bullish IREN and was made more comfortable by Microsoft accepting Horizon 1, but argues the $110B-by-January-2028 scenario at 25x earnings requires $4.4B annual net income, leaving only $3.85B for depreciation, net interest and taxes against an $8.25B EBITDA illustration, which he calls tight given GPU investment.
Against the $8.25B EBITDA illustration, that leaves $3.85B for depreciation, net interest and taxes combined.
This is a direct disagreement on IREN's terminal economics: even bulls flag that GPU depreciation and dilution could absorb most of the EBITDA, capping per-share value.
Watch Confirm via IREN's disclosed GPU useful life, financing costs, and share count trajectory; invalidation would be longer useful lives or cheaper financing than assumed.
Source →Existing-model demand alone justifies years of AI infrastructure
The author argues better models are not needed to justify years of buildout because deploying today's models already exceeds buildable compute, citing Anthropic's 80x vs 10x planned growth, OpenAI's stated compute shortage, Google's 7x token growth, HBM sold out at all three makers, TrendForce 50-60% HBM shipment growth, Nvidia's ~70% revenue growth guide called supply-constrained against customer forecasts near 100%, and Goldman's view that only about half of scheduled US capacity will be built on time.
NVIDIA, the biggest buyer of it, expects its revenue to grow about 70% next year and calls that outlook 'supply-constrained',
Shifts the AI capex debate from model-quality risk to physical supply constraints, supporting memory, power and data-center exposure even if frontier progress stalls.
Watch HBM shipment growth and Nvidia's next guidance versus the ~70% supply-constrained figure and the ~100% customer forecast.
Source →