Microsoft Is Significantly Behind Its Own Targets For AI Chip Deployment: Report

A Guardian investigation published this week has cast fresh doubt on the pact of enterprise AI adoption, finding that Microsoft is significantly behind its own targets for AI chip deployment across its global data centers.

Despite committing billions annually to expanding data center capacity, Microsoft sources told the outlet the company’s total chip count has barely moved over the past year.

Internal documents cited in the report put the current figure at 2.2 million AI chips installed, less than half of what some analysts had projected for this stage of the rollout, as demand continues to outstrip supply from the small group of manufacturers that control the global AI chip market.

This discrepancy raises questions about Microsoft’s competitive position in the AI race, particularly whether it can meet the capacity commitments tied to its data center investments.

The news follows a month of mounting local backlash against data center construction, as communities grow concerned with potential strain on water reserves particularly. In the U.S., data centers are now a firm political issue, with New York State putting a halt on the construction of any facility rated at 50 megawatts or more.

What began as localized opposition is slowly but surely scaling into a national, cross-partisan movement that could slow the country’s race for AI supremacy.

A more conservative approach to compute?

While this shift is taking place against the backdrop of capacity challenges, enterprises have also been caught short by bill shock this year as AI companies began charging by usage-based tokens rather than leveraging seat-based pricing.

For large enterprises, keeping control of innovation and R&D budgets is only part of the story. A broader capacity crunch could disrupt AI initiatives regardless of budget, and push usage costs higher still.

Some providers are already adjusting course. Ness Digital Engineering, for one, is focused on modernizing legacy enterprise systems for AI rather than building new infrastructure from scratch. IBM, on its end, has introduced a modernization tool aimed at the same problem. Budget discipline is becoming as central to enterprise AI strategy as infrastructure itself.

“Real value shows up when you run AI like you run any other part of the business,” said Brandon Tobman, CEO of Get Covered, a company developing AI software solutions for the property insurance sector, while in conversation with Forbes.

“Someone owns it, there’s a defined outcome you’re measuring against and it actually lives inside your workflow instead of sitting off to the side as a science project,” he explained, adding that this does not necessarily require the entire overhauling of technology budgets, but does call for reallocating resources toward efforts that show measurable operational impact.

Shifts in AI leadership signal change

The future course of AI progress is also being shaken up by major leadership changes this week after Google announced a leadership overhaul of its AI division. Demis Hassabis, a name synonymous with the early AI industry, has stepped down as chief executive at Google DeepMind, although he will retain a position as chair of the company.

The announcement, however, doesn’t mean that Hassabis is keen on an early retirement. Instead, his efforts are refocused on exploring singularity, despite the controversy around artificial general intelligence (AGI) both inside and outside the AI industry.

“It’s critical that we collectively get the next steps right to ensure this all goes well for humanity and we usher in an incredible new age of discovery and wonder,” Hassabis said in a statement last week.

The comments come amid a string of security-related incidents across major AI companies this year, including OpenAI’s Hugging Face hacking scandal and a ban on Anthropic’s powerful Mythos model, further underscoring the stakes involved.

Google has also seen key developments, including the exit of veteran engineer Jeff Dean, who has left to launch a startup called Discovery Loop. Separately, Sonata Software named Hariprasad Rebala as Chief AI Officer, tasked with leading AI-native delivery for enterprise clients.

In a company statement, Rebala said that AI alone will not determine which companies succeed, but rather “the ability to translate AI into enterprise velocity.”

Speed is also just one part of the journey that engineering leaders are taking to ensure value is derived from AI. A The Economist analysis published this August noted that much of the value generated by AI accrues quietly to individual employees through small, scattered time savings, making it harder to quantify at an organizational level.

One thing that is clear from these developments is that the actions of AI leaders today will shape the future of the industry significantly, from the frontier models through the success of business transformation initiatives.

Demand for dynamic memory causes 400% surge in costs

Rising demand for data center capacity is also straining memory supply chains. New research from J.P. Morgan Global Research projects that prices for dynamic random access memory (DRAM), a core component of data-intensive AI workloads, will rise more than 400% by the end of 2026, in just a two-year window.

Given how deeply electronics are embedded in daily life, that increase is expected to reach consumers directly, as manufacturers pass costs through to everyday products like smartphones, smartwatches, and laptops.

With AI now embedded into everyday software programs and enterprise operations worldwide, the question is no longer about selling the potential of AI, either. It’s about finding a way to manage resources and their impact on everything from the billion-dollar data center buildout to the price in the storefront tomorrow.

As demand for AI resources grows, leaders and experts will need to place a renewed focus on not only value, but speed, ease and quality to ensure that AI investment continues to generate returns as compute and memory costs climb.