Barclays Finds AI Is Now Essential for Institutions, But Andreessen Questions Grid Readiness

A new Barclays AI survey shows institutional investors now use AI across daily research and risk work. The findings give real-world weight to Marc Andreessen’s warning that energy and cooling will decide how far it can scale.

The survey polled 410 fixed-income investors across North America, Europe, the Middle East and Asia. It found AI has moved past testing and into everyday use, though humans still make the final calls.

Barclays AI Survey Shows Research Leads Adoption

Research is the leading use. About 52% of long-only managers and asset owners use AI mainly for research, Barclays found. Another 44% of hedge funds lean on it to process market data.

Hedge funds are the heaviest users. Some 72% report using AI daily, compared with 49% of long-only managers and 38% of asset owners. That gap tracks broader signs of strong institutional AI demand.

By contrast, AI stays on the sidelines in trading and execution. Most respondents see only minor impact there, and they ranked data security as the top barrier to wider use.

Few expect the shift to cost jobs. Only 7% foresee meaningful staff cuts, while most predict higher output and steady headcount.

Andreessen Ties AI Growth To Energy And Cooling

Marc Andreessen, cofounder of venture firm Andreessen Horowitz, tied AI’s future to physical limits in a recent post. The framing extends his long-running case for energy abundance.

The AI:AC Hypothesis. In the future, in each country, the amount of AI will be proportional to the amount of AC. And vice versa,” Marc Andreessen said in a post.

Follow us on X to get the latest news as it happens

His point is about power and heat. AI servers draw heavy power, and cooling them draws more, feeding a steep rise in AI electricity consumption.

The IEA expects data center demand to more than double by 2030, to about 945 terawatt hours. That is close to Japan’s total power use today.

The strain falls hardest on the United States. There, data centers may soon use more power than US aluminum, steel and cement production combined, per the IEA. Regions with cheap, reliable power will attract the most AI.

What The Pairing Means For Investors

Together, the two threads tell one story. Barclays shows demand for AI is already here. Andreessen flags the energy and cooling limits that will sort the winners from the rest.

Institutional money is funding both sides. The hyperscalers building that capacity include Microsoft, Amazon, Alphabet and Meta.

The four have set out a combined $725 billion in 2026 capital guidance, up 77% on this year.

The grid keeping pace may shape the next phase of AI energy consumption debates and the returns that follow.