The Compute Triad: Korea's Ascent and the End of US Monopoly
The narrative that the US controls the AI hardware stack via Nvidia's design and TSMC's manufacturing is officially dead. Samsung Electronics' $200 billion partnership with Broadcom, announced alongside the commercial shipment of HBM4, signals a structural shift in global semiconductor power. This isn't just a business deal; it's a geopolitical realignment.
Samsung is no longer the also-ran chasing TSMC's process nodes. By securing massive custom AI accelerator contracts and dominating HBM, Samsung is leveraging its vertical integration to offer what fabless giants cannot: a bundled silicon-memory solution. With SK Hynix reporting record Q2 profits and holding a 57% share of the HBM market, Korea has become the indispensable swing state of the AI era. Nvidia's $1 billion investment in Naver and expanded accord with SK Group confirms that even the US chip king needs Korean capital and capacity to scale.
The blind spot: Markets are pricing Korea as a passive beneficiary. They are wrong. Samsung's rise creates a bifurcated supply chain. If Washington pushes too hard on export controls, Seoul may quietly pivot toward deeper integration with Asian markets, as evidenced by the massive capital flows into CXMT, China's DRAM champion. CXMT's poised debut as the biggest China-listed firm demonstrates Beijing's strategy: use state capital to IPO champions, normalize their global presence, and erode sanctions efficacy over time. The "Silicon Curtain" is becoming a mesh network where Korea and China are finding ways to bypass US friction.
The Open Source Threat
Meanwhile, Silicon Valley is fracturing over restrictions on Chinese AI. The summary of the day's debate is stark: China is closing the gap not through proprietary breakthroughs alone, but through a ruthless embrace of open source and massive compute deployment. DeepSeek's R1 model topping US app charts before pausing its funding round due to viral backlash highlights the volatility of this race. Beijing's approach is utilitarian; it doesn't need to own the IP if it can deploy the models.
My call: The US policy of tech containment is failing. By 2027, China will have a functional, sovereign AI stack powered by open-source models running on domestically produced memory (CXMT) and logic chips. The gap is narrowing faster than policymakers admit. Samsung's dominance in HBM means the US is effectively arming both sides of this divide, creating a strategic vulnerability that Washington will struggle to manage.
The Macro Hammer: Oil, Bonds, and the AI Liquidity Trap
While tech headlines scream about valuations, the macro data is flashing red. Global bonds are reeling as an oil surge rekindles inflation fears, delivering brutal losses to investors who bet the worst was over. This is the story that matters. The AI capex boom is financed by debt, and debt is suddenly becoming expensive again.
BlackRock's soft demand for a bond sale tied to a Meta data center project is the canary in the coal mine. We are seeing an "AI debt sell-off" that contradicts the equity frenzy. Investors are buying AI stocks but rejecting AI bonds. This divergence suggests the market believes in AI's revenue potential but fears the cost of capital and execution risk.
The contradiction: The market is pricing in infinite AI growth while ignoring the resurgence of energy-driven inflation. Oil prices are rising due to supply constraints and geopolitical risk, which pushes rates higher. Higher rates crush the valuation of long-duration assets like AI infrastructure. If oil holds above current levels, we will see a repricing of the entire AI trade. Companies relying on leveraged capex, particularly non-hyperscalers, will face a liquidity crunch.
Forward look: Expect credit spreads to widen for mid-tier AI firms in Q3 2026. The "AI trade" is no longer uniform. Hyperscalers with cash flow can weather the storm; everyone else needs to refinance. BlackRock's struggle signals that the bond market is de-risking from AI capex bets. If inflation sticks, the Fed may be forced to pause rate cuts, delivering a double whammy to tech valuations.
Valuation Whiplash: Security vs. Consumer AI
The valuation landscape is splitting into two distinct universes: security AI and consumer AI. Anduril's move toward a $100 billion valuation and the Boring Company's $20 billion mark reflect a market that rewards defense and infrastructure plays with certainty of demand. Anduril benefits from counter-cyclical government spending, insulated from consumer slowdowns.
Conversely, Morgan Stanley's note on SpaceX suggests the market is questioning the AI revenue contribution from satellite networks. The slide in SpaceX implies that AI value is not materializing as quickly as projected in some segments. Rivian's lawsuit to recover Trump-era tariffs adds another layer of risk: the trade war is costing automakers billions, and legal battles will only escalate supply chain friction.
The irony: As DeepSeek pauses funding after viral success, we see the paradox of the open-source era. Success can trigger regulatory or public backlash faster than revenue can scale. Meanwhile, Microchip's acquisition of Israeli AI firm Hailo shows US integrators are buying capability rather than building it, accelerating the consolidation of edge AI.
My opinion: The market is overpaying for defense AI and underestimating the durability of consumer AI applications. Meta's Seller app and Anthropic's Opus 5 (offering near-Fable 5 capabilities at half the price) show that efficiency gains are real. But the valuation disconnect will persist until oil prices cool. Until then, investors should rotate from leveraged capex plays to cash-rich integrators and defense tech.
Regional Shifts: Capital Flight to Asia
Capital is rotating rapidly. Investors are pouring money into India's AI and SaaS startups, as well as Southeast Asia's health and SaaS sectors. Hong Kong is easing listing thresholds to attract IPOs, desperate to reclaim its status as a fundraising hub. This is a direct response to US-China tensions. Global capital is seeking jurisdictions with growth potential but lower geopolitical risk.
The profitability of Grab's Superbank in SEA versus the struggles of other digital banks highlights the winner-take-all dynamics in emerging markets. Only integrated ecosystems can survive. Meanwhile, the fraud lawsuit against Futu's founder underscores the governance risks in Chinese brokers trying to access global markets. Trust is becoming a currency.
The bottom line: The AI infrastructure war is shifting from US-dominated fabless design to an integrated Korean-Chinese duopoly, while a resurgence in oil prices is exposing a fragility in AI debt markets that bulls are ignoring. Samsung's $200B Broadcom deal cements Korea as the new center of gravity for compute, but macro headwinds are intensifying. Investors must pivot from pure growth narratives to strategic resilience: favor integrated supply chains, defense AI, and cash-rich hyperscalers, while avoiding highly leveraged capex plays. The era of easy AI money is over; the age of geopolitical and macro discipline has begun.