As the AI revolution matures, the most compelling opportunities may not lie in who builds the next ChatGPT, but in who powers it
Three years after the launch of ChatGPT, generative artificial intelligence (Gen AI) has become one of the most transformative forces in global technology markets. According to Hilary Frisch, Portfolio Manager at ClearBridge Investments, the expansion of Gen AI across both consumer and enterprise sectors has triggered a powerful investment cycle, particularly for companies that build the infrastructure underpinning this technological revolution.
“We’re still in the early innings of a once-in-a-decade capital expenditure cycle,” said Frisch. “Cloud hyperscalers continue to dominate AI investment, but the real opportunity for investors lies in the ‘picks and shovels’ industries that enable large language models to run at scale.”
Cloud leaders Microsoft, Amazon, Alphabet and Meta Platforms are projected to spend a combined US$378 billion in 2025- a 65% increase from 2024 – as they expand their data centre and AI capabilities. Frisch noted that this surge in spending, funded largely through internal cash flows rather than debt, “demonstrates confidence in AI as a long-term evolution in computing, not a speculative trend.”
Emerging players such as Oracle, CoreWeave, and several privately held cloud providers are also committing billions to AI infrastructure, adding further momentum to the capex supercycle.
Frisch noted the critical role of cloud infrastructure and data software companies in enabling AI workloads. As enterprises move from experimental in-house AI systems to proven, scalable platforms, demand for sophisticated infrastructure software is surging.
Oracle, for example, has transformed into a formidable fourth hyperscaler. Its Oracle Cloud Infrastructure powers major AI initiatives such as the U.S. government’s Stargate AI project and partnerships with OpenAI, xAI, and Meta. “Oracle’s evolution shows how tailored infrastructure built for AI workloads can open new growth avenues,” said Frisch.
Meanwhile, data warehousing firms like Snowflake and Databricks are thriving as companies look to unify siloed data for AI training and analytics. “Quality data is the fuel for large language models,” Frisch explained. “These platforms are increasingly indispensable to the AI ecosystem.”
Frisch also pointed to growth in monitoring and observability software, a niche sector helping enterprises manage complex technology stacks. Firms such as Datadog and Dynatrace are poised to benefit as organisations scale up AI workloads and demand deeper visibility into performance and security.
The unprecedented compute demands of Gen AI are reshaping the semiconductor landscape. “Inference models like ChatGPT require specialised chips that traditional processors simply can’t handle,” said Frisch. “This has sparked a renaissance in custom silicon development.”
While Nvidia remains the dominant player in AI GPUs, companies such as Broadcom, Marvell Technology, and ARM Holdings are carving out lucrative niches by designing application-specific integrated circuits (ASICs) tailored for hyperscaler needs.
“Broadcom’s partnerships with Google, Meta, and OpenAI highlight a smart strategy—collaborating rather than competing with the hyperscalers,” Frisch noted. “We expect AI chip revenues to accelerate significantly as these data centres scale.”
Frisch added that networking and connectivity solutions like Marvell’s high-speed ethernet switches and optical interconnects are becoming just as critical as processing power. “The next bottleneck in AI computing isn’t the chip but how fast data can move between them,” she said.
Despite the hype, Frisch believes the Gen AI boom is only beginning. “We’re witnessing a secular growth trend across infrastructure software and semiconductors,” she said. “As hyperscalers sustain record-level capex, companies building the AI backbone- those that make the systems run- are positioned for durable earnings and cash flow growth.”
“In short, as the AI revolution matures, the most compelling opportunities may not lie in who builds the next ChatGPT, but in who powers it.”
