The wave of corporate spending on generative artificial intelligence is not, in Goldman Sachs’s view, an overexcited bubble dismantled by reality; it is the start of a multi-year investment cycle that could underpin significant economic value. In a Wednesday note, Goldman analysts wrote that “the enormous economic value promised by generative AI justifies the current investment in AI infrastructure,” and that “overall levels of AI investment appear sustainable as long as companies expect that investment today will generate outsized returns over the long run.” The bank estimates roughly $300 billion in AI-related spending in 2025.

That figure matters because it anchors a market debate. Critics have likened the recent run-up in AI-linked equities and capital expenditure to past tech excesses, but Goldman’s analysis frames current spending as driven by expectations of durable productivity gains rather than mere speculation. From the perspective of corporate finance, capital projects that expand data-center capacity, buy specialized chips, or retool software platforms are meaningful only if they raise future margins or lower costs. Goldman’s baseline is that many firms — particularly cloud providers, software incumbents and financial firms automating knowledge work — expect such outcomes.

The market implications are immediate. Sustained capex at this scale supports demand for high-performance semiconductors, cloud infrastructure, and advanced services from systems integrators. That in turn can justify higher valuations for suppliers that capture a durable revenue stream. Conversely, concentrations of spending raise the risk of winner-take-most dynamics: a handful of chip and cloud vendors may reap most revenue, while many downstream users could underdeliver on promised productivity gains.

Goldman’s stance also reframes policy priorities. If AI capex is a long-lived structural shift, public policy should focus less on short-term market stabilization and more on managing transitions: workforce retraining, competition policy to prevent monopolistic capture of critical infrastructure, and standards for data governance and security. Policymakers face trade-offs between incentivizing private investment — via tax treatment of intangible investment and R&D credits — and ensuring those incentives do not entrench market power or leave workers behind.

The note does not dismiss risks. Returns on complex digital capital can be lumpy and uncertain; firms that overinvest relative to their ability to integrate AI into products and processes risk write-downs. There are also macro feedbacks: if AI spending disproportionately inflates certain sectors, it could exacerbate capital misallocation and financial fragility when sentiment shifts.

Longer term, Goldman’s outlook hinges on diffusion. The most significant macro payoff will come if generative AI moves beyond a narrow set of use cases and materially boosts productivity across manufacturing, health care, professional services and public administration. That transition can take years if not decades, which is why the bank treats current large-scale capex as an investment in prospective returns rather than a speculative froth to be popped.

For investors and policymakers, the watchwords are measurement and patience. Trackable metrics — productivity gains, cost savings per application, and realized revenue lift — will determine whether the $300 billion estimate buys an enduring economic transformation or simply funds a reshuffling of winners and losers.