Best AI Stocks to Watch After Big Tech Earnings: 4 Infrastructure Bottlenecks

AI stocks to watch: HBM memory, electrical power grid, server CPU and optical networking in data centers
DAYHOOK INVESTING / THE STORY IN 10 SECONDSBest AI Stocks to Watch After Big Tech Earnings: 4 Infrastructure Bottlenecks
1Five key takeaways
21. What Big Tech earnings reveal about AI spending
32. Google Clouds 103% growth: demand, TPUs and profitability
THE TAKEAWAY

What are the best AI stocks to buy after Big Tech earnings? The more useful question may be which suppliers can turn a surge in hyperscaler spending into durabl

What are the best AI stocks to buy after Big Tech earnings? The more useful question may be which suppliers can turn a surge in hyperscaler spending into durable free cash flow. Microsoft’s, Alphabet’s, Amazon’s and Meta’s latest filings point to four infrastructure categories worth monitoring: AI memory, electrical power, server CPUs and optical networking. These are research themes, not a blanket buy recommendation.

Important fact-check: The source summary describes Google Cloud growth of 63%. Alphabet’s official Q3 2026 release instead reports 103% year-over-year Google Cloud revenue growth to $30.7 billion. This article uses the reported 103% figure. The spending numbers below are total corporate property-and-equipment capital expenditures, not a verified AI-only subset.

THE BIG PICTURE

Follow the evidence, compare opportunities and keep risks in view.
AI INFRASTRUCTURE · CONCEPT ILLUSTRATION

Where Big Tech Spending Flows

01

Cloud capex

02

Memory & CPUs

03

Power & optics

Follow orders, margins and free cash flow — not spending headlines alone.

🔎 Five key takeaways

  • Google Cloud’s reported 103% growth underscores AI infrastructure demand, but revenue and profitability must be evaluated together.
  • Four hyperscalers spent approximately $202 billion on property and equipment in the September 2026 quarter; these are company-wide figures.
  • HBM and DRAM, grid connections, server CPUs and optical links have different supply and pricing dynamics.
  • A bottleneck can create supplier pricing power but also cause project delays and higher buyer costs.
  • Balance-sheet quality, backlog conversion and free cash flow matter more than AI-related branding.

📈 1. What Big Tech earnings reveal about AI spending

Alphabet reported Q3 2026 CapEx of $51.4 billion and maintained its full-year 2026 guidance of $195–205 billion. The company said approximately 65% of technical infrastructure investment in Q3 was for servers and networking and about 35% for data centers. Microsoft had previously guided to approximately $190 billion of calendar-year 2026 CapEx, including about $25 billion attributed to higher component pricing. These statements support continued investment, but they do not prove that every AI supplier will earn superior returns.

Visual 1 · Q3 2026 reported capital expenditure, approximate
Amazon

$77.3B total CapEx

Alphabet

$51.4B total CapEx

Microsoft

$39.4B total CapEx

Meta

$33.9B total CapEx

Dayhook Investing editorial illustration; conceptual, not a measured market share or forecast.

The combined figure is about $202 billion in the quarter, based on cash-flow statement line items compiled from filings. Definitions can vary and some spending supports non-AI businesses. Large spending is evidence of procurement activity, not necessarily a guarantee of future investment returns.

💡 2. Google Cloud’s 103% growth: demand, TPUs and profitability

Alphabet reported Google Cloud Q3 2026 revenue of $30.7 billion, up 103%, with cloud operating income of approximately $11.1 billion and an operating margin of 36.2%. Management identified AI infrastructure, AI solutions and TPU systems sales as contributors. TPU system sales also affect comparisons with periods when cloud revenue was dominated by traditional services.

The video summary highlights Gemini adoption, Anthropic-related demand and custom TPU usage. Those themes are plausible drivers of the broader ecosystem, but specific customer contributions should not be inferred without company disclosure. Investors should track how much growth translates into recurring revenue, operating profit and free cash flow.

Visual 2 · The AI infrastructure demand chain
AI applications

Gemini, enterprise copilots and model APIs

Compute demand

Accelerators, server CPUs and memory

Facility capacity

Electricity, cooling and data halls

Connectivity

High-bandwidth optical and network links

Dayhook Investing editorial illustration; conceptual, not a measured market share or forecast.

🤖 3. Four AI infrastructure sectors to watch

🔹 Sector A · HBM, DRAM and enterprise storage

HBM supplies bandwidth for AI accelerators, while conventional server DRAM, NAND and enterprise SSDs serve the surrounding system. Memory suppliers can benefit from tight high-performance supply, but commodity memory cycles and new capacity can reverse prices quickly. The source video claims per-user DRAM needs may rise fivefold in five years; this is a scenario rather than a verified industry forecast.

Representative names: SK Hynix, Samsung Electronics and Micron. Watch: HBM qualification, average selling prices, inventory, supply additions and customer concentration.

⚡ Sector B · Power grids, transformers and electrical distribution

Electricity access and grid interconnection can be harder to scale than server orders. Transformers, switchgear and power-management systems are necessary to energize data center sites. However, long backlogs are only valuable if projects receive permits, are delivered and convert into cash.

Representative names: Eaton, HD Hyundai Electric, Hyosung Heavy Industries and LS Electric. Bloom Energy is a separate on-site power-generation theme with different economics and technology risk.

🔹 Sector C · Server CPUs and platform architecture

Even GPU-heavy AI systems require host processors, memory management and orchestration. AMD and Intel compete in server CPUs while Arm licenses architectures used in some custom server designs. CPU supply is not interchangeable with advanced packaging for accelerators: TSMC CoWoS constraints can affect certain chip packages, but do not establish a universal shortage of all server CPUs.

Representative names: AMD, Intel and Arm. Watch: server market share, performance per watt, software compatibility, manufacturing yields and hyperscaler custom-silicon adoption.

🔹 Sector D · Optical networking, transceivers and photonics

As AI clusters scale, bandwidth and power-efficient interconnects become more important. Optical transceivers and laser components can benefit from higher port speeds and additional links. The source summary’s claim that interconnect demand scales universally with the square of GPU count (N²) is not a general engineering law; topology, network oversubscription and workload determine actual requirements.

Representative names: Coherent and Lumentum. Watch: 800G/1.6T adoption, customer qualifications, laser yields, product mix and pricing.

Visual 3 · Four bottlenecks and the metric that matters
Memory

HBM yield, DRAM pricing and inventory

Electrical power

Transformer lead time and backlog conversion

Server CPUs

Share, platform efficiency and margins

Optical links

Port speed, shipments and gross margin

Dayhook Investing editorial illustration; conceptual, not a measured market share or forecast.

Original editorial illustration for this article's topic; conceptual, not a financial data chart
DayHook Investing original illustration · Conceptual visualization, not measured market data.

⚡ 4. Dark horse: power semiconductors

Power semiconductors, including silicon carbide and gallium nitride devices in suitable applications, can improve conversion efficiency. But the investment case depends on system design wins, qualification, cost and thermal trade-offs—not every AI facility will automatically switch to the same technology.

📈 5. How to track real AI demand during volatile markets

OpenRouter activity: public usage dashboards can offer a directional view of third-party model traffic. They do not measure all private, first-party or enterprise AI usage, so the video’s roughly 10x growth claim should not be extrapolated to the entire market without consistent methodology.

GPU rental rates and utilization: rental quotes, contract terms, chip generations and observed availability can indicate supply-demand conditions. Public spot prices are imperfect proxies for private long-term hyperscaler contracts.

Visual 4 · The investor evidence ladder
Demand signal

Tokens, cloud revenue and deployed workloads

Supply signal

Lead times, capacity and utilization

Economics

Pricing, margins and cash conversion

Valuation

Free cash flow, leverage and expectations

Dayhook Investing editorial illustration; conceptual, not a measured market share or forecast.

DAYHOOK INVESTING / GLOBAL WATCHLISTTOP PICKS 155 US · 5 INDIA · 5 SOUTH KOREA / THEMATIC RESEARCH

USUNITED STATES

  1. Micron Technology (MU)
  2. Eaton (ETN)
  3. AMD (AMD)
  4. Coherent (COHR)
  5. Lumentum (LITE)

ININDIA

  1. ABB India (ABB.NS)
  2. Bharti Airtel (BHARTIARTL.NS)
  3. Larsen & Toubro (LT.NS)
  4. Tata Consultancy Services (TCS.NS)
  5. Tech Mahindra (TECHM.NS)

KRSOUTH KOREA

  1. SK Hynix (000660.KS)
  2. Samsung Electronics (005930.KS)
  3. HD Hyundai Electric (267260.KS)
  4. Hyosung Heavy Industries (298040.KS)
  5. LS Electric (010120.KS)
THEMATIC RESEARCH WATCHLIST · NOT PERSONALIZED INVESTMENT ADVICE

🔴 Bottom Line: Top Picks 15

This watchlist includes five US, five Indian and five South Korean listed companies to retain Dayhook Investing’s global research format. Indian companies are primarily indirect cloud, connectivity, engineering or enterprise-AI exposures rather than direct HBM, CPU or optical suppliers. Selection is thematic, not a claim that all fifteen are attractive at today’s valuations.

Micron Technology (MU)US

Sector: Memory & storage

Why · AI memory demand

  • HBM and server DRAM demand expands with AI accelerator deployments.
  • Enterprise SSDs support data-intensive workloads and retrieval.
  • Memory pricing may benefit from tight high-end supply.
Risk · Memory cycles, HBM execution and

  • Memory cycles can reverse quickly as capacity expands.
  • HBM qualification and yields determine mix and profitability.
  • Large customer spending changes can trigger inventory corrections.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

Eaton (ETN)US

Sector: Power & grid

Why · Grid connections

  • Switchgear and power management help connect data centers.
  • Electrical distribution upgrades can extend beyond individual AI projects.
  • Installed equipment creates recurring service opportunities.
Risk · Permitting delays

  • Grid permitting and project delays defer deliveries.
  • Input costs and execution can reduce margins.
  • High valuations may anticipate years of demand.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

AMD (AMD)US

Sector: Server CPUs

Why · Server CPU share

  • EPYC CPUs are used in enterprise and hyperscale servers.
  • AI clusters still require host CPUs alongside accelerators.
  • GPU and CPU portfolios allow broader platform sales.
Risk · Intense competition

  • Intel and Arm-based alternatives intensify competition.
  • Cloud customer concentration makes orders volatile.
  • Supply constraints and product transitions can limit share gains.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

Coherent (COHR)US

Sector: Optical networking

Why · Cluster interconnects

  • Optical components help connect dense compute clusters.
  • Higher bandwidth demand supports advanced transceivers and lasers.
  • Data center connectivity is an adjacent AI infrastructure spend.
Risk · Optics cycles, qualification yields

  • Optical product cycles can be volatile.
  • Customer qualification and manufacturing yields matter.
  • Rapid technology shifts can pressure margins.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

Lumentum (LITE)US

Sector: Optical networking

Why · Photonics growth

  • High-speed optical components support AI cluster links.
  • Advanced lasers and photonics address bandwidth upgrades.
  • New optical architectures may expand content per connection.
Risk · Customer concentration

  • Customer concentration and qualification delays pose risks.
  • Optics inventory cycles can be sharp.
  • Technology competition may erode pricing power.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

ABB India (ABB.NS)India
Why · Grid electrification for AI capacity

  • Electrical distribution and automation are relevant to data-center power constraints.
  • Domestic grid and industrial electrification spending can support equipment demand.
  • A diversified power-equipment business provides a different exposure than cloud platforms.
Risk · Industrial orders and valuation

  • Data centers are only one part of company-wide demand.
  • Project timing and industrial cycles can pressure orders.
  • Rich valuations may already price in infrastructure growth.
Bharti Airtel (BHARTIARTL.NS)India

Sector: Network infrastructure

Why · Enterprise networks

  • Enterprise connectivity links cloud and data center demand.
  • Data center partnerships provide indirect infrastructure exposure.
  • Network upgrades support AI traffic growth.
Risk · Indirect AI earnings

  • AI demand is not a direct earnings proxy.
  • Telecom regulation and capex affect cash flow.
  • Pricing competition can pressure returns.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

Larsen & Toubro (LT.NS)India

Sector: Power & construction

Why · Grid engineering

  • Engineering expertise supports data center and electrical projects.
  • Large infrastructure execution can benefit from grid investment.
  • Diversified backlog provides exposure beyond AI.
Risk · Project overruns

  • Execution delays and cost overruns hurt margins.
  • Working capital intensity can affect cash conversion.
  • AI-specific revenue may be a small share of the group.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

Tata Consultancy Services (TCS.NS)India

Sector: AI implementation

Why · Enterprise AI integration

  • Enterprise AI adoption requires integration and data modernization.
  • Cloud partnerships support implementation work.
  • Recurring client relationships may support long projects.
Risk · Automation disruption

  • Automation could reduce legacy service demand.
  • Discretionary IT budgets remain cyclical.
  • AI infrastructure hardware exposure is indirect.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

Tech Mahindra (TECHM.NS)India
Why · Network and cloud systems integration

  • Cloud and network modernization connect enterprise infrastructure to AI deployments.
  • Systems integration can help businesses deploy new computing workflows.
  • Telecom and enterprise clients provide a different demand mix from pure-play chip suppliers.
Risk · Telecom budgets and execution

  • Telecom customer spending can be cyclical.
  • AI projects may not translate directly into incremental profit.
  • Margin recovery depends on execution and staffing costs.
SK Hynix (000660.KS)Korea

Sector: Memory & HBM

Why · HBM bandwidth, premium packaging and

  • HBM supports bandwidth-intensive AI accelerators.
  • Advanced memory packaging can command premium pricing.
  • AI inference may broaden long-term memory requirements.
Risk · Memory cyclicality

  • Memory pricing and demand remain cyclical.
  • HBM capacity expansion can change supply dynamics.
  • Qualification and concentration risks remain significant.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

Samsung Electronics (005930.KS)Korea

Sector: Memory & storage

Why · AI server memory

  • DRAM, HBM and enterprise SSDs address AI server needs.
  • Broad semiconductor portfolio spans multiple infrastructure layers.
  • Manufacturing scale can support product diversification.
Risk · HBM execution, oversupply and foundry

  • HBM yields and customer qualification affect competitiveness.
  • Memory oversupply can pressure profitability.
  • Foundry and consumer cycles complicate the investment thesis.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

HD Hyundai Electric (267260.KS)Korea

Sector: Power & grid

Why · Transformer shortages

  • High-voltage transformers support power delivery to data centers.
  • Grid modernization can sustain orders beyond AI.
  • Long lead times may provide order visibility.
Risk · Permitting delays

  • Utility and permitting delays can shift deliveries.
  • Capacity expansion could eventually ease shortages.
  • Input costs and foreign exchange affect margins.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

Hyosung Heavy Industries (298040.KS)Korea

Sector: Power & grid

Why · Grid interconnection

  • Power transformers address grid interconnection bottlenecks.
  • Global transmission spending broadens demand sources.
  • Specialized manufacturing supports high-voltage projects.
Risk · Contract execution

  • Contract execution and working capital create risk.
  • Lead-time normalization could pressure pricing.
  • Grid spending can be sensitive to regulation.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

LS Electric (010120.KS)Korea

Sector: Power distribution

Why · Switchgear demand

  • Switchgear and distribution systems support new facilities.
  • Power automation improves reliability and efficiency.
  • Grid investment may benefit electrical equipment makers.
Risk · Diversified exposure

  • AI-specific demand is only part of the business.
  • Price competition and material costs can pressure margins.
  • Project timing affects order conversion.

Opens a new tab with external research results; verify the exchange, coverage and date. Dayhook does not publish an unsupported target.

🤖 What could break the AI infrastructure bull case?

  • Hyperscalers reduce capital spending after disappointing returns on deployed compute.
  • Memory manufacturers add capacity faster than demand and trigger a pricing reversal.
  • Grid permits, transmission access or financing delay data center projects.
  • Custom accelerators or network architectures change the supplier mix.
  • High stock valuations leave little room for earnings misses despite solid industry growth.

🔹 Frequently asked questions

📈 What are the best AI infrastructure stocks to buy?

There is no universal best stock independent of price and risk. Investors can screen memory, power, CPU and optics companies for competitive position, balance-sheet strength and valuation.

💡 Why did Google Cloud grow so quickly in Q3 2026?

Alphabet cited AI infrastructure, AI solutions, core cloud demand and TPU system sales. It reported 103% year-over-year segment revenue growth.

📈 Are AI memory stocks safer than GPU stocks?

No. Memory suppliers face cyclical pricing and manufacturing investment, while GPU suppliers face platform competition and customer concentration. Compare cash flow and valuations.

🔹 Is transformer demand guaranteed to rise?

Not guaranteed. Power grid expansion and data center projects create demand, but permitting, utility budgets and order timing can shift.

🔹 Do optical links scale as N squared with GPUs?

Not as a universal rule. Actual optical port requirements depend on network architecture, topology, oversubscription and workload.

💡 Why do the target price buttons open a search?

They open current external analyst-consensus research in a new tab rather than presenting unverified targets. Confirm the exchange, analyst coverage and publication date.

📈 Related Dayhook Investing research

🔹 Sources and editorial methodology

Editorial note: This article adapts a user-supplied earnings commentary summary. Official disclosures take precedence when they conflict with the source summary. No measured Google Keyword Planner search volumes or SEO competition scores were available, so keyword selection is editorial rather than a claim of verified ranking difficulty.

Disclaimer: For informational and educational purposes only. This is not personalized financial advice or a recommendation to purchase securities. Research consensus targets independently and consider the possibility of capital loss.


Discover more from Dayhook Investing

Subscribe to get the latest posts sent to your email.