AI Market 2025 Year-End Review

Big Tech’s AI infrastructure surge

“The last two years have been defined by a hyperscaler arms race, with Microsoft (NASDAQ:MSFT), Alphabet (NASDAQ:GOOGL) and others spending multi-tens of billions per quarter on GPUs, servers, fiber and data centers,” Nicholas Mersch, portfolio manager at Purpose Investments, told the Investing News Network (INN).

As Amazon (NASDAQ:AMZN), Microsoft, Alphabet and Meta Platforms (NASDAQ:META) pledged investments worth over US$300 billion on AI data centers, GPUs and custom chips, power emerged as the binding constraint.

“AI data centers are now running into hard power limits. Individual campuses are pushing past 1 gigawatt, (and) utilities in key regions are scrambling to add generation and transmission,” said Mersch. “Forecasts for US data center electricity demand have been revised sharply higher as AI loads are expected to roughly triple into 2030.”

Data centers, which accounted for roughly 4 percent of US electricity in 2023, are now projected to consume up to 12 percent of the nation’s total power by 2028. “Even Google has acknowledged that serving capacity needs to double roughly every six months," explained Mersch. "The constraint is shifting from capital to kilowatts, which pulls utilities, nuclear operators and grid planners into the center of the AI story.”

Companies pursued energy-efficient designs, including optimized chips to curb surging data center electricity demand, and renewable power investments.

Investors shift AI funding efforts

The technology sector continued to account for a large portion of S&P 500 (INDEXSP:.INX) earnings growth in 2025, buoyed by innovation in semiconductors, AI and cloud services. Funding shifted toward AI-native firms with proven revenue, favoring consumer-facing applications over pure infrastructure.

Venture capital poured into AI-native companies showing annual recurring revenue growth and profitability paths, with AI accounting for a staggering 63.3 percent of all venture capital deal value in 2025, as per PitchBook data.

The tech industry saw remarkable M&A activity this past year, highlighted by strategic acquisitions focused on enhancing AI expertise, data center capacity and scalable platforms.

Deal making reflected a shift in buyer priorities toward securing talent and infrastructure critical for the AI era.

Late-stage mega rounds dominated, led by Big Tech’s direct investment in labs, signaling consolidation over early experimentation. According to Bloomberg, Jane Street Group’s trading revenue for the third quarter surged by US$830 million, primarily driven by successful strategic investments in private AI companies, with Anthropic-related gains reportedly accounting for about 12 percent of Jane Street's US$6.83 billion total.

Against that backdrop, AI stock valuations remain elevated, with some pockets of the market showing signs of speculative valuation risks, prompting bubble questions.

Semiconductor trends reinforce AI momentum

John Murillo, B2BROKER’s chief business officer, dismissed comparisons of AI hype to the dot-com boom.

“The prevalent part of the AI market is occupied by the tech giants like Google and Microsoft. These companies have proved for years their ability to generate profits among any instabilities … moreover, their pricing is more than fair, because it is constantly reevaluated by professional auditors," he commented.

“The reasons behind every situation are completely different. In the case of dot-coms, everyone was investing just to invest; it didn’t matter what exactly to choose and some of the projects didn’t have a solid foundation. With AI, it’s not like this. The technology proves its worthiness every day, and it has already swept away many junior analysts.”

Semiconductor dynamics reinforce the staying power of AI.

“Underneath the headline GPU names, the chip stack is quietly diversifying,” Mersch observed, adding that high-bandwidth memory (HBM) — the fast RAM powering AI GPUs — has become the new profit center, with suppliers racing to HBM3E and upcoming HBM4 versions as demand explodes. This dynamic is showing up in pricing, with Counterpoint Research forecasting a 50 percent price rise for memory modules through the second quarter of next year.

What’s more, on a call to clients on November 26, Dell Technologies (NYSE:DELL) COO Jeff Clarke noted that he has never seen costs rise as quickly as they are now, citing tighter supply and rising costs for DRAM (including HBM for AI and PC chips), hard drives and NAND flash memory across all products.

Clarke said the impact will inevitably reach customers, leading Dell to “consider all options,” including repricing devices.

Verticalization: The next phase of the AI race

“Data center investment remains elevated and heavily concentrated in a handful of cloud platforms,” said Mersch.

“Over the next 12 to 24 months, the narrative likely shifts from who can build fastest to who can drive the highest revenue and margin per dollar of AI infrastructure. This is where verticalization matters. The companies that can capture the full stack, from silicon to applications, look like they will win. Top pick here is Google, followed by Microsoft.”

As the industry moves past the hype, the new race is on. The winners of the AI era will be those who can optimize power, scale and profitability from silicon to end-user applications.