|
Feb 17, 2026
|
|
LONG
|
Campling argues the key differentiator in AI is data ownership. "Alphabet... owns the data. Microsoft doesn't own the data [mainly through OpenAI JV]." In the long-term AI race, margins and capability will accrue to those who control the feedstock (data). Google's ownership of DeepMind and Search data gives it a structural margin advantage over Microsoft, which relies on a partnership model. LONG GOOG/GOOGL (Data Owner) vs. WATCH MSFT (Data Renter). Regulatory breakup of Google; OpenAI achieves AGI faster than Google despite data disadvantage. |
Bloomberg Markets
India Seeks Role in AI Future As Modi Hosts F...
|
|
Feb 17, 2026
|
|
LONG
|
"I think at this point there are three or four platforms that had the scale that could be an alternative source... And, of course, the Pentagon has had some deep relationships with of the big tech companies for a number of years." Anthropic's ethical hesitation ("safeguards against basically mass surveillance") creates a friction point in securing government defense contracts. The Pentagon's demand for AI capability is urgent. If Anthropic stalls, the capital and contracts will flow to the "alternative sources" — the legacy Hyperscalers (Microsoft, Google, Amazon) who already possess the required scale and deep, existing security clearances/relationships with the DoD. LONG US Big Tech as the default beneficiary of defense AI spending when ethical pure-plays (like Anthropic) decline participation. Regulatory pressure on Big Tech regarding AI safety could increase; Anthropic might eventually concede to secure revenue. |
Bloomberg Markets
Anthropic in Disagreement With Pentagon Over ...
|
|
Feb 17, 2026
|
|
SHORT
|
Dassault Systemes and Siemens sold off sharply (Dassault ~8-10%). UBS issued a report questioning the defensibility of industrial design software. Schindler (elevator co) reportedly dropped Dassault to build internal tools. Investors previously viewed complex industrial software (Digital Twins) as a "moat." The inference is that Generative AI allows companies to build these tools in-house cheaper and faster, destroying the pricing power and retention of legacy industrial software vendors. SHORT/AVOID Industrial Software incumbents. The sell-off is an overreaction to a single client loss (Schindler); AI integration might eventually aid these platforms. |
Bloomberg Markets
Anthropic’s Pentagon Talks Snag, Pound Falls ...
|
|
Feb 16, 2026
|
|
AVOID
|
"Samsung is also saying they will be impacted negatively because they make smartphones, PCs and TVs... inflate prices on everything from laptops to cars." While Samsung makes chips, their massive exposure to consumer electronics means they are a net victim of rising component costs. Higher memory costs crush margins on hardware or kill demand if passed to consumers. AVOID hardware manufacturers with high memory exposure who lack the pricing power of the hyperscalers. Samsung's memory division profits outweighing their consumer electronics losses. |
Bloomberg Markets
Memory Chip Shortage is Global Crisis in the ...
|
|
Feb 16, 2026
|
|
LONG
|
"With the latest AI chips they used 10 times or 6 times the amount of memory... it takes 3-5 years to build a new memory fabrication plant, and that is creating a bottleneck." Demand is exponential (AI) while supply is inelastic (multi-year build times). This creates a classic super-cycle for memory manufacturers where pricing power shifts entirely to the producers. LONG pure-play memory manufacturers and semiconductor equipment suppliers. Global recession dampening AI capex; rapid resolution of supply chain issues (unlikely given the 3-5 year lead time). |
Bloomberg Markets
Memory Chip Shortage is Global Crisis in the ...
|
|
Feb 16, 2026
|
|
LONG
|
AI chips (specifically Nvidia's Rubin) use 8-10x more memory than H100s. Hyperscaler capex is $600B, but memory supply is constrained. Spot prices are estimated to be up 60% QoQ. You cannot build a fabrication plant in 3 months. The supply-demand imbalance is structural and worsening. This grants immense pricing power to the memory oligopoly (Micron, SK Hynix, Samsung). South Korea (EWY) is the geographic proxy for this trade. LONG Memory Producers and Korean Equities. Global recession crushing demand for consumer electronics (phones/PCs) which these companies also rely on. |
Bloomberg Markets
'Shared Values' discussed in Munich; RAM Conc...
|
|
Feb 16, 2026
|
|
AVOID
|
Memory spot pricing is skyrocketing (up 60%). For hardware manufacturers (Dell, HP, Apple), memory is a major input cost (COGS). They face a "double whammy": rising component costs and a consumer potentially weakened by inflation. They cannot easily pass these costs on without killing demand. AVOID Hardware Integrators due to margin compression. Companies successfully pass costs to consumers or AI-PC demand outweighs price increases. |
Bloomberg Markets
'Shared Values' discussed in Munich; RAM Conc...
|
|
Feb 12, 2026
|
|
AVOID
|
"Yesterday... it was real estate stocks... CBRE, a good example, was hit double digits... fear that there could be [an AI agent] launched." The market is currently in a "shoot first, ask questions later" mode regarding AI disruption. Service-heavy sectors (like real estate brokerage) are being repriced for existential risk, regardless of current earnings. Avoid until sentiment stabilizes. Earnings call dispels fears and stock snaps back. |
Bloomberg Markets
Stocks Climb; Nuveen to Buy Schroders; Anthro...
|
|
Feb 12, 2026
|
|
SHORT
|
"Commercial real estate stocks that were hit hard... potential for models being released [that disrupt the sector]." AI is viewed as a deflationary force for office demand (fewer humans needed). The market is pricing in terminal value risk for CRE similar to how it priced Kodak. Short/Avoid Commercial Real Estate. Return-to-office mandates stabilizing occupancy. |
Bloomberg Markets
Nuveen to Buy Schroders in £10B Deal | The Pu...
|