HSBC says 2026 flips the AI profit script to software—yet Nvidia just posted 75% data-center growth. The takeaway: don’t rotate away from silicon; own both ends of the stack.
The New Thesis: Software Eats 2026
HSBC tech analyst Stephen Bersey told clients the AI winners of 2026 will be companies monetizing code, not chips. His logic: enterprise software is sticky, renewal rates top 90%, and AI workloads are finally becoming production-grade reliable inside Microsoft 365, Google Cloud, and ServiceNow. Translation—customers will pay premium seat-prices for generative tools that shave hours off legal, finance, and coding tasks.
Bersey cites Microsoft’s $34 billion quarterly productivity revenue (up 16%) as proof that corporate wallets open for bullet-proof software faster than they’ll overhaul servers. The bank’s model sees software margins punching past 80% once AI co-pilots scale, dwarfing the 55–60% gross margin ceiling for high-end GPUs.
Hardware Isn’t Dead—It’s Just Front-Loaded
Two days before HSBC’s call, Nvidia posted fiscal Q4 2026 data-center revenue of $62.3 billion, a 75% year-over-year burst that beat consensus by $4.6 billion. Management guided Q1 to $78 billion—77% growth—implying annual run-rate demand north of $300 billion. CFO Colette Kress told analysts hyperscalers are still under-supplied on H100 and the new B300 platforms, pushing 2026 cap-ex budgets to between $3 trillion and $4 trillion industry-wide.
Meta Platforms reinforced that message hours later, signing a >$100 billion procurement pact with AMD for 6 GW of custom AI silicon—enough silicon to power roughly 10 million next-gen GPUs over five years. The multi-fab order gives AMD an estimated 10% stake in Meta’s future compute stack and shatters the myth that hyperscaler spend is plateauing.
Valuation Check: Where the Odds Sit Today
- Nvidia trades at 28× next-12-month (NTM) earnings, a 20% discount to its three-year median despite 60% average EPS growth.
- Alphabet fetches 19× NTM earnings, the cheapest among megacaps, while cloud revenue accelerated 48% to $17.7 billion last quarter, fastest since 2022.
- Microsoft commands 30× NTM earnings, but free-cash-flow margin is 35% and Copilot attach rates reached 35% of eligible seats in Q2.
Straddle the Stack, Don’t Time the Handoff
History shows every compute wave—PC, internet, cloud—rewards both component builders and application vendors. Intel dominated the late-1990s while Microsoft hit record highs; Salesforce surged post-2010 while rack-dense server stocks also outperformed. Investors who bifurcated won twice.
For 2026, that means:
- Keep core Nvidia or AMD for continued accelerator refresh cycles.
- Add software bellwethers levered to seat pricing—Microsoft, ServiceNow, Snowflake—where Gen-AI upsell is still under 10% penetration.
- Hedge with two mid-cap picks levered directly to AI software spend: Datadog (observability for AI workloads) and CrowdStrike (AI-driven security).
Red Flags to Watch
- Supply Glut: If Nvidia inventory days climb above 120, pricing power erodes.
- Cap-Ex Pause: Should the cloud five (Microsoft, Amazon, Alphabet, Meta, Oracle) guide 2027 spend below $200 billion cumulatively, trim chip beta quickly.
- Regulation: EU AI Act liability rules debut Q3 2026—enterprise software adoption could decelerate if per-seat compliance costs exceed $8/month by 2027.
The AI profit pie is expanding faster than investors can draw slices. The traders moving “all-in” on software risk missing another year of 70% GPU growth; the die-hard semis-only crowd could give back gains when software billings inflect. Balance the barbell—own the silicon that trains, own the software that sells—today, not tomorrow.
Stay ahead of every earnings inflection with rapid-fire analysis—bookmark onlytrustedinfo.com for the fastest stock-moving insight on the web.
