Skip to content
TBSI.

Start here

  • LLM LeaderboardCompare frontier models on intelligence, price and speedTool
  • What is the best superintelligence?The best frontier model for each jobModels
  • Frontier Gap ReportHow far apart the leading labs areReport
  • MethodologyHow we verify sources, data and claimsAbout

AI and Superintelligence Stocks to Watch in 2027: A Six-Layer Map

Seventeen stocks across six layers of the AI stack, each with the one 2027 signal that matters

By TheBestSuperintelligence Editorial TeamUpdated

◈ Executive SummaryEvidence · Primary Source

Which AI and superintelligence stocks should you watch in 2027, layer by layer?

Watch the AI stack layer by layer: compute (Nvidia, AMD, Broadcom), memory (Micron, SK Hynix), fabs and tools (TSMC, ASML, Arm), power (GE Vernova, Constellation, Vistra), cloud (Amazon, Alphabet, Meta) and software (ServiceNow, Salesforce, Palantir). Each has one 2027 signal. Information, not a recommendation.

Educational research, not investment advice. Prices move and sources can be wrong: check company filings and consider a licensed adviser before you buy.

☰On this page (8 sections)
  1. 01.The 2027 AI stock map: six layers to watch
  2. 02.Why 2027 is the year AI spending has to pay back
  3. 03.Will AI capex in 2027 keep rising?
  4. 04.The 2027 AI stocks, layer by layer
  5. 05.The 2027 superintelligence IPO window
  6. 06.What could go wrong for AI stocks in 2027?
  7. 07.Limits of this guide
  8. 08.Sources & Further Reading

The 2027 AI stock map: six layers to watch

  1. 1
    ComputeSignal: Rubin's share of revenue; Helios rack volume; OpenAI custom accelerators
    Nvidia, AMD, Broadcom2027 question: Does custom silicon take share from GPUs?
  2. 2
    MemorySignal: contract pricing on long-term agreements
    Micron, SK Hynix2027 question: How long do peak margins last?
  3. 3
    Fabs, tools and IPSignal: ASML order intake
    TSMC, ASML, Arm2027 question: Can capacity grow fast enough?
  4. 4
    PowerSignal: Crane restart date; new hyperscaler power agreements
    GE Vernova, Constellation, Vistra2027 question: Can electricity keep up with data centers?
  5. 5
    CloudSignal: cloud revenue growth versus capex growth
    Amazon, Alphabet, Meta2027 question: Does capex finally earn a return?
  6. 6
    Agentic softwareSignal: agent revenue disclosed as its own line
    ServiceNow, Salesforce, Palantir2027 question: Do agents become line-item revenue?

Numbers are stack layers, not rankings. Select a layer to jump to its stocks.

Not financial advice. This article is for information only and is not a recommendation to buy or sell any security. All figures are company-reported and dated. Forward-looking statements are company guidance, not forecasts by this publication.

Why 2027 is the year AI spending has to pay back

2026 was the year AI spending went vertical. 2027 is the year it has to pay back, and the year the labs chasing superintelligence may finally list. Think of AI as a stack where money flows down and constraints flow up: hyperscalers spend, and chipmakers, memory suppliers, foundries and power companies collect. Every layer enters 2027 supply-constrained.

This is the same ladder of constraints we track in the top 10 chip bottlenecks and the data center bottleneck.

Will AI capex in 2027 keep rising?

Guidance says yes. The question is whether revenue follows.

Company 2026 capex 2027 guidance
Alphabet $195–205B "Significantly" higher
Amazon About $220B cash capex "Meaningfully more"
Meta $135–145B (floor raised) "Invest aggressively"
Four hyperscalers combined $720–740B guided (EconoTimes) Citi: $801B for Alphabet, Amazon and Meta

The 2027 AI stocks, layer by layer

Compute (NVDA, AMD, AVGO)

Who supplies the chips in 2027?

Nvidia (NVDA) enters 2027 with Rubin in production and the deepest software moat in AI. Fiscal Q2 2027 data center revenue was $89.0B (+117%), and the company says Vera Rubin is in full production (Nvidia). Its CUDA ecosystem is the real lock-in (analysis).

AMD (AMD) has become a real second source. Q3 2026 revenue was $13.4B (+45%) and data center revenue $8.5B (+97%), led by the Instinct MI450 launch. It guided Q4 to about $17B, and named OpenAI, Anthropic, Meta, Microsoft and OCI as running workloads on its first Helios rack systems (AMD).

Broadcom (AVGO) is the custom-accelerator layer for labs that want to own their silicon. AI semiconductor revenue was $16.7B in fiscal Q3 (+221%), with $21.7B guided for Q4 (Broadcom).

FocusNVDA · AMD · AVGO

Memory (MU, SK Hynix)

How long does the supercycle last?

Micron (MU)'s revenue roughly tripled in a year. Fiscal 2026 revenue was $133.19B, against $37.38B in fiscal 2025, and fiscal Q4 gross margin was 86.8% (Micron).

SK Hynix (000660.KS) is the HBM leader, and it has pre-sold the future. Q3 2026 revenue was ₩104.6T (+328%) with an operating margin near 80%. It says HBM4 overtakes HBM3E in Q4 2026, and HBM and DRAM are sold out through 2029 (Korea Times).

FocusMU · SK Hynix (000660.KS)

Fabs, tools and IP (TSM, ASML, ARM)

Can supply catch up?

TSMC (TSM) expects about 44% US-dollar revenue growth in 2026 and is spending $60–64B on capex. High-performance computing was 68% of Q3 revenue (Vested).

ASML (ASML) says it cannot build enough machines. Q3 2026 net sales were €11.7B, full-year guidance is €44–45B, and management said 2027–28 supply "will not fully meet the demand," with capacity expansion beyond a planned 30% increase for 2028 under study (ASML).

Arm (ARM) is turning data center CPUs into its growth engine. Fiscal Q2 2027 revenue was $1.443B (+27%), and data center royalties doubled year over year (Arm).

FocusTSM · ASML · ARM

Power (GEV, CEG, VST)

Can electricity keep up with data centers? In 2027, a data center without a grid connection is an expensive warehouse. Power is now slower to secure than chips (analysis).

GE Vernova (GEV) reported Q3 2026 orders of $29B (+96% organically) and backlog of $197B, which it expects to pass $200B in Q4 (GE Vernova).

Constellation Energy (CEG) signed another 920 MW of long-term nuclear power agreements in Q2 2026 and targets the Crane restart for 2027 (Constellation). Background: nuclear's second act.

Vistra (VST) closed its Cogentrix acquisition on October 30, 2026, adding about 5.5 GW of gas capacity and bringing its fleet to nearly 50 GW (Enerdata).

FocusGEV · CEG · VST

Cloud (AMZN, GOOGL, META)

Does capex pay back?

Amazon (AMZN): AWS grew 39% to $46.0B in Q3 2026 with about $220B of 2026 cash capex (Amazon); management said 2027 capex will be "meaningfully more" (TechCrunch).

Alphabet (GOOGL): Google Cloud grew 103% in Q3; 2026 capex guidance is $195–205B and 2027 capex is expected to increase "significantly" (Google; transcript).

Meta (META) raised the floor of its 2026 capex guidance to $135–145B and says it will "invest aggressively" in 2027 (Meta). It also told investors returns from Meta Superintelligence Labs "will take longer" than its core business (transcript).

FocusAMZN · GOOGL · META

Agentic software (NOW, CRM, PLTR)

Do agents become revenue?

ServiceNow (NOW) raised its 2026 AI net new ACV target from $1B to $1.7B and reported agentic AI consumption up more than 13x (ServiceNow).

Salesforce (CRM) reported Agentforce ARR above $1.5B (+240%) (Salesforce).

Palantir (PLTR) grew US commercial revenue 148% in Q3 2026 (XTB).

FocusNOW · CRM · PLTR

The 2027 superintelligence IPO window

2027 could be the first year investors can buy a superintelligence lab directly.

Lab Status (Nov 6, 2026) Last private valuation
Anthropic Public S-1 filed Oct 26, 2026 (Nasdaq: ANT); timing uncertain $965B (May 2026)
OpenAI Confidential filing Jun 2026; "IPO ready" in 2027 per its president $852B (Mar 2026)
xAI (in SpaceX) Listed via the SpaceX IPO, Jun 2026 ~$1.75T SpaceX at IPO (Reuters)
Safe Superintelligence Private; Nvidia investing ~$5B $32B (Apr 2025)

Which public stocks benefit most if these list? See best superintelligence stocks to buy in 2027 for the exposure ranking, frontier labs compared for the lab picture, and AI stocks to watch through the end of 2026 for the dated catalysts before year-end.

What could go wrong for AI stocks in 2027?

  • Payback gap. If capex keeps outgrowing cloud revenue, valuations compress, the risk covered in AI valuation reality.
  • Cycle turn in memory. Margins near 80–87% are historically peak-cycle.
  • IPO supply. Trillion-dollar listings pull capital from existing AI holdings.
  • Price deflation. Cheaper tokens raise usage but squeeze unit economics (inference cost paradox).
  • Policy. Export controls, EU AI Act enforcement and grid permitting.

Final thought. In 2027 the AI trade splits: suppliers with sold-out capacity versus spenders who must prove payback. The superintelligence IPOs will force a public price on the whole stack.

Limits of this guide

This does not predict prices. 2027 statements are company guidance or third-party forecasts as of November 6, 2026. Some backlog figures come from earnings-call summaries rather than filings. SK Hynix figures are from press reports of its results. Private valuations are last-round marks and may not reflect public-market pricing.

Sources & Further Reading

Claim Source Date
Citi's $801B combined 2027 capex forecast Tradingpedia, reporting Citi (we did not read Citi.s note) 2026-07-16
Four hyperscalers' $720–740B 2026 capex EconoTimes 2026-10-30
SK Hynix Q3 2026, sold out through 2029 Korea Times 2026-10-28
GE Vernova Q3 2026, backlog $197B GE Vernova 2026-10-28
AWS backlog $578B AlphaStreet, Amazon Q3 2026 call 2026-10 (day not stated)
OpenAI "IPO ready" in 2027 TechCrunch 2026-11-05
Nvidia Q2 FY27, data center revenue $89.0B Nvidia, via Finviz 2026-08-26
AMD Q3 2026, MI450 and Helios AMD 2026-11 (day not stated)
Broadcom AI revenue $16.7B, Q4 guide $21.7B Broadcom, via Finviz 2026-09-02
Micron fiscal 2026 revenue $133.19B Micron, via Pulse 2.0 2026-09-30
TSMC Q3 2026, capex $60–64B Vested 2026-10-15
ASML Q3 2026, 2027–28 supply ASML, via MFN 2026-10 (day not stated)
Arm fiscal Q2 2027 Arm 2026-11 (day not stated)
Constellation: 920 MW of agreements, Crane restart Constellation 2026-08 (day not stated)
Vistra closes Cogentrix acquisition Enerdata 2026-10-30
AWS +39%, 2026 capex ~$220B Amazon 2026-10-29
Amazon 2027 capex "meaningfully more" TechCrunch 2026-10-29
Google Cloud +103%; Alphabet capex guidance Google; Q3 transcript 2026-10-28
Meta capex guidance and "invest aggressively" Meta Q3 2026 results 2026-10 (day not stated)
Meta: MSL returns "will take longer" Meta Q3 2026 call transcript 2026-10-21
ServiceNow AI net new ACV target ServiceNow 2026-10 (day not stated)
Salesforce Agentforce ARR Salesforce 2026 fiscal Q2 FY27 (day not stated)
Palantir Q3 2026 US commercial revenue XTB 2026-11-02
Anthropic Series H at $965B Anthropic 2026-05-28
OpenAI at $852B Built In 2026-04 (day not stated)
SpaceX IPO at ~$1.75T Reuters via BusinessDay 2026-06-11

Frequently asked questions

What AI stocks should I watch in 2027?

Watch the AI stack layer by layer: compute (Nvidia, AMD, Broadcom), memory (Micron, SK Hynix), fabs and tools (TSMC, ASML, Arm), power (GE Vernova, Constellation, Vistra), cloud (Amazon, Alphabet, Meta) and software (ServiceNow, Salesforce, Palantir). Each has one 2027 signal. Information, not a recommendation.

Will AI capex keep rising in 2027?

Guidance points up. Alphabet expects 2027 capex to increase significantly, Amazon expects it to be meaningfully higher than about $220 billion in 2026, and Meta says it will invest aggressively. Citi forecast in July 2026 that Alphabet, Amazon and Meta would spend a combined $801 billion in 2027.

Will OpenAI or Anthropic list on the stock market in 2027?

Anthropic filed a public S-1 on October 26, 2026 to list on Nasdaq as ANT; its timing became uncertain after the November 5, 2026 selloff. OpenAI filed confidentially in June 2026, and its president said on November 5, 2026 that it could be IPO ready in 2027.

What is the biggest bottleneck for AI stocks in 2027?

Supply, not demand. SK Hynix says its HBM and DRAM are sold out through 2029, ASML says 2027–28 tool supply will not fully meet demand, and GE Vernova expects at least 130 GW of gas equipment under contract by the end of 2026.

How we sourced this. Draft supplied by the site owner from the sources below; not independently re-read by our editors. Facts current as of 2026-11-06. Methodology