Maintain core equity exposure to NVIDIA (NVDA), which continues to dominate the AI landscape by recycling over $70 billion in annual free cash flow back into its ecosystem to guarantee compute demand.

For high-upside exposure to the physical data center buildout, look to memory leaders Samsung Electronics and SK Hynix via the South Korean Equities (KOSPI) market, while bracing for cyclical volatility.

Pre-IPO and crossover investors should track Anthropic for a near-term foundation model market debut and Stripe for surging AI transaction volume ahead of OpenAI's anticipated 2027 public listing.

Consider allocating capital toward proprietary AI training data providers like Mercor, which serve as high-growth infrastructure bottlenecks for frontier models.

Conversely, avoid deploying capital into pure AI Customer Support tools, Humanoid Robotics, and early-stage Defense Tech startups, as these niches face heavy commoditization, slow enterprise procurement, and high deployment hurdles.

NVIDIA (NVDA)

• NVIDIA is aggressively investing its cash flow into every layer of the AI ecosystem, including paying $6 billion to license Poolside's model factory, investing $1 billion into Poolside at a $12 billion valuation, and moving 109 engineers to bolster its Nemotron model.
• The company is participating heavily in funding rounds for core infrastructure and application layers, including Mercor's $20 billion round and Perplexity's $30 billion round.
• NVIDIA's strategy mirrors vendor financing: allocating an estimated $70 billion to $75 billion in annual free cash flow back into its customers and ecosystem partners, effectively ensuring sustained demand for its compute chips.
• Championing open-source and open-weight models serves as an economic complement for NVIDIA, as open models direct enterprise spending toward hardware and compute rather than proprietary closed-model software margins.

Takeaways

• Bullish ecosystem control: NVIDIA's dominance is reinforced by recycling massive free cash flows into the broader ecosystem, making its platform essential across hardware, foundation models, and applications.
• Watch vendor financing risks: While effective in high-growth cycles, heavy reliance on funding one's own customer base carries systemic risk if end-user corporate demand fails to generate sufficient return on investment.

OpenAI (Private)

• OpenAI CFO Sarah Friar confirmed internal targets for the company to go public in 2027, driven by capital intensity and pressure from rivals.
• Growth metrics showed signs of deceleration in early periods, with quarter-over-quarter revenue growth tracking at 18% (from ~$5 billion to ~$6 billion+ run-rate levels), though internal reports indicate an aggressive push to reaccelerate.
• OpenAI faces intensifying competition from open-weight models and rival foundation models, putting pressure on its valuation multiple and market position as an enterprise platform.
• The consumer model remains financially challenging; heavy power users can consume $8,000 to $12,000 worth of compute tokens on subscriptions that cost only $200, creating negative unit economics unless subsidized by high-margin enterprise products or future advertising.

Takeaways

• IPO on the horizon: Public investors should prepare for a massive liquidity event by 2027, though pricing and valuation multiples will heavily depend on whether OpenAI maintains clear market leadership over open-weight alternatives.
• Enterprise monetization is critical: Long-term profitability requires shifting focus from consumer ChatGPT subscriptions to high-margin developer, coding, and enterprise infrastructure workflows.

Anthropic (Private)

• Anthropic is reported to be planning an initial public offering in the near term, outpacing competitors to the public markets.
• The company is operating at an estimated $60 billion annualized run-rate mid-year, with long-term internal projections targeting $200 billion in GAAP revenue by 2028.
• Anthropic has demonstrated significant margin expansion over the past year, improving gross margins from -91% to positive 30%.
• Claude has established dominant market share in software engineering and enterprise coding workflows, which currently represent the highest-ROI segment in generative AI.

Takeaways

• First-mover advantage in public AI markets: Anthropic's earlier timeline for an IPO could make it the premier pure-play public foundation model stock, setting the benchmark multiple for later offerings.
• Coding remains the primary revenue driver: Software development remains the clearest monetization avenue for generative models, giving Anthropic a strong competitive moat in enterprise adoption.

Stripe (Private)

• Stripe demonstrated strong reacceleration, growing revenue 41% year-over-year with billings up 71%, driven heavily by its role as the default payment infrastructure for AI agents and services.
• The company acquired OpenRouter to position itself as a manager and allocator of AI compute and intelligence routing.
• Demonstrating exceptional financial health, Stripe has actively reduced its total share count over the past several years through private share buybacks while self-funding acquisitions.

Takeaways

• Prime crossover asset: Stripe represents one of the strongest pre-IPO software assets, combining stable, cash-generative core payment processing with pure upside exposure to agentic AI transactions.
• Benchmark for late-stage SaaS: High-scale private companies growing above 40% while buying back shares present severe competition for investor capital once they enter public markets.

Mercor & AI Data Providers (Private)

• Mercor is scaling rapidly, crossing $2.5 billion ARR and raising new capital at a $20 billion valuation led by General Catalyst, with participation from NVIDIA.
• High-quality training and labeling data providers are becoming critical infrastructure, with speculation that top data providers could represent up to 5% to 10% of the frontier model market capitalization.
• Data service businesses face valuation debates over whether they should be priced at premium AI software multiples or discounted services multiples due to gross margin profiles.

Takeaways

• Bottleneck infrastructure: As base models mature, proprietary and specialized data providers represent a vital layer of the supply chain, commanding high revenue growth even if gross margins resemble technical services rather than pure software.

Semiconductor Memory Sector & South Korean Equities (KOSPI)

• The South Korean market (KOSPI) acts as a high-beta proxy for AI hardware demand, up 56.46% year-to-date, driven primarily by memory suppliers such as Samsung Electronics and SK Hynix.
• The sector has experienced sharp volatility—surging to 9,000 before dropping to 5,600 and rebounding—reflecting shifting sentiment around hardware capital expenditure cycles.
• Unprecedented gross margin expansion across high-bandwidth memory (HBM) and specialized semiconductor components continues to fuel strong performance.

Takeaways

• High volatility, high leverage on AI capex: Memory manufacturers offer direct exposure to physical data center builds, but investors should expect severe quarterly drawdowns tied to supply-chain adjustments and macro sentiment.

Overvalued & Challenged AI Sectors

• Customer Support & CX: Pure customer service AI tools face heavy commoditization as standard AI agents absorb customer interaction workflows, and large enterprises opt to build in-house systems.
• Humanoid Robotics: Despite significant venture capital interest, general-purpose humanoid robots face major economic and deployment hurdles compared to specialized, single-purpose industrial robotics.
• Defense Tech Startups: While strategically critical, early-stage single-product defense tech firms face severe procurement bottlenecks with the Pentagon, likely leading to consolidation into major prime contractor platforms rather than independent multi-billion-dollar exits.
• AI Roll-ups of Professional Services: Attempting to acquire traditional legal or accounting partnerships and transform them into tech-enabled high-margin software platforms presents complex structural and operational challenges with unclear exit potential.

Takeaways

• Avoid narrow workflow tools: Prioritize companies building deep infrastructure, proprietary data assets, or foundational platforms over thin application layers in customer service or tech-enabled services roll-ups.