Tech stocks today: Meta, CoreWeave agree to $21 billion deal

CoreWeave and Meta struck a $21 billion AI compute deal through December 2032, and Meta also launched its Muse Spark model (Meta stock +9% intraday). Amazon reports AWS AI revenue run-rate >$15B and a chip business run-rate of $20B (estimated ~$50B standalone), while imposing a 3.5% fuel/logistics surcharge for third-party sellers. Broadcom agreed to produce Google TPUs and supply Anthropic ~3.5 GW from 2027 (Anthropic revenue run-rate >$30B), and SpaceX is targeting a $2 trillion IPO valuation. Ongoing US-Iran hostilities and explicit threats to data centers (e.g., OpenAI’s Stargate) keep geopolitical risk elevated and volatility possible for tech shares.

Analysis

The market is re-pricing a persistent structural shortage in AI-grade compute and memory as a multi-year cash flow tailwind rather than a one-off cycle. That means margins will shift up the stack toward component suppliers (memory, interconnect, rack-level ASICs) and specialist integrators who can guarantee low-latency, high-density deployments; expect revenue acceleration to show up in reported gross margins for those suppliers over the next 2-4 quarters.

Hyperscalers pushing vertically into custom silicon is a real competitive pressure but not an instantaneous moat-breaker — systems integration, software stacks, and interconnect scale take 12–24 months to replicate at meaningful cost parity. The transitional window creates a sweet spot: incumbents with ecosystem lock-in retain pricing power near-term while well-capitalized entrants can win share in niche, high-margin workloads if they scale global capacity and partnerships.

Geopolitical risk to physical compute assets is an underpriced convex tail risk that raises expected capex and insurance costs for operators with concentrated regional exposure. In practice this will accelerate localization of data centers to allied jurisdictions and push buying toward suppliers able to deliver geographically diversified capacity, altering TTM capex plans over the next 6–18 months.

Consensus is treating Big Tech as a single theme; the next 6–12 months will deliver dispersion. Earnings cadence, third-party capacity announcements, and memory-price trajectories will drive meaningful divergence — active selection and tactical hedging will beat passive exposure in this regime.

Key Decisions for Investors

• Buy NVDA 12-month call spread (debit) to express continued GPU demand: size 2–4% portfolio, target 40–60% return if enterprise deployments accelerate; max loss = premium paid (controls downside vs outright long).
• Long MU (Micron) equity or 9–12 month LEAPs to capture memflation tailwinds: target +50% over 6–12 months if data-center memory ASPs stay elevated; set stop at 25% downside to limit cyclicity risk.
• Long AVGO (Broadcom) 9–18 months to play networking/connectivity capture from rack-level AI deployments — convex upside if hyperscalers accelerate private-rail investments; hedge 30–50% notional with short NVDA single-month calls around major launch dates to fund cost.
• Small-cap exposure to specialist AI cloud integrators (size 0.5–1% portfolio in CRWV or equivalent) via long equity or long-dated calls to capture outsized contract wins; keep position size low and take profits on 30–50% moves due to concentration risk.
• Buy 3–6 month tail protection (OTM puts) on a basket of cloud/hyperscaler exposure (AMZN, MSFT, GOOGL) sized to cover 10–15% portfolio drawdown — rationale: protects against a geopolitical/data-center outage shock that would compress multiple names simultaneously.