AI Stocks Since ChatGPT

AI chip designers create the GPUs and specialized processors that power artificial intelligence. NVIDIA dominates with 80%+ market share in AI training chips, while AMD and Broadcom compete for the rest. These companies design chips but outsource manufacturing to foundries like TSMC.

Why they're winning: Training large language models like GPT-4 and Claude requires thousands of high-end GPUs running for months. A single AI training run can cost $100 million+ in compute. As AI models grow larger, demand for specialized AI chips continues to surge.

Chip equipment companies make the machines that manufacture semiconductors. They're the "picks and shovels" of the AI gold rush. Lam Research, Applied Materials, KLA, and Teradyne supply the lithography, etching, and testing equipment that foundries need to build AI chips.

Why they're winning: Every new AI chip fab requires billions in equipment. As TSMC, Samsung, and Intel race to build advanced chip factories, equipment makers benefit from both AI demand and government subsidies like the CHIPS Act.

Beyond AI-specific chips, the broader semiconductor industry benefits from AI infrastructure buildout. Super Micro Computer (SMCI) makes AI servers, while ON Semi, Microchip, NXP, Analog Devices, and Texas Instruments provide power management and analog chips needed in every electronic device.

Why they matter: AI data centers need more than GPUs—they require power delivery chips, networking components, and server infrastructure. These companies often have diversified revenue but benefit from AI tailwinds.

AI models need massive amounts of memory. High Bandwidth Memory (HBM) from Micron and SK Hynix sits directly on AI chips, providing the data throughput that training requires. Western Digital and Seagate provide storage for the petabytes of training data.

Why memory is crucial: A single NVIDIA H100 GPU uses 80GB of HBM3 memory. Training GPT-4 reportedly required 25,000 GPUs—meaning 2 petabytes of HBM alone. Memory companies have pricing power as HBM demand far exceeds supply.

AI data centers consume enormous amounts of electricity. A single AI training run can use as much power as a small city. Vistra, Constellation Energy, and NRG Energy own power plants—including nuclear facilities—that can provide the reliable, large-scale electricity that AI requires.

The surprise AI winners: Power utilities were unexpected beneficiaries of the AI boom. Microsoft signed a deal to restart Three Mile Island for AI power. Google and Amazon are buying nuclear power for data centers. These utilities have surged as investors recognize AI's insatiable energy appetite.

The tech giants are racing to integrate AI across their products. Microsoft has Copilot and a $13B investment in OpenAI. Google has Gemini and dominates AI research. Meta built Llama, the leading open-source model. Amazon powers AI infrastructure through AWS.

AI as a moat: These companies have the data, compute, and distribution to deploy AI at scale. Microsoft embeds Copilot in Office. Google adds AI to Search. Meta uses AI for content recommendation. Their returns are strong but more muted than pure-play AI chip stocks.

Enterprise software companies are adding AI features to justify premium pricing. Salesforce has Einstein GPT for CRM. ServiceNow uses AI for IT automation. Adobe added Firefly for generative images. Palantir pivoted to become an "AI company" with its AIP platform.

The AI upsell opportunity: These companies can charge more for AI-enhanced products. Salesforce added AI tiers. Adobe raised Creative Cloud prices. The question is whether AI features drive enough value to justify higher costs—or whether AI commoditizes their core products.

Cloud infrastructure companies help businesses run and monitor their applications. Datadog provides observability for AI workloads. Cloudflare powers content delivery and edge computing. Snowflake offers data warehousing that feeds AI models.

AI infrastructure layer: These companies benefit from increased cloud spending but face competition from hyperscalers. Datadog sees more logs from AI applications. Snowflake stores training data. Cloudflare accelerates AI inference at the edge.

Cybersecurity companies use AI to detect threats faster than human analysts. CrowdStrike uses AI for endpoint protection. Zscaler secures cloud access with AI. Palo Alto Networks integrates AI across its security platform.

AI arms race in security: As attackers use AI to create more sophisticated threats, defenders need AI to keep up. These companies benefit from both increased attack surfaces (more cloud, more AI apps) and the need for AI-powered defense.

Some SaaS companies face disruption from AI. Zoom's video meetings could be replaced by AI avatars and async video. DocuSign's e-signatures seem commoditized. Workday's HR software faces AI-powered competitors. These pandemic darlings have struggled in the AI era.

The disruption question: Can AI build a better Zoom meeting? Will contracts need signatures if AI agents negotiate? These companies are adding AI features, but investors worry their core products could become commodities or be replaced entirely by AI-native alternatives.