Goldman Sachs: The Next Wave of AI Investment Shifts to Factories, Mines, and Power Grids in the Physical Economy
Taylor Wilson
Goldman Sachs projects $7.6 trillion in global AI infrastructure spending from 2026 to 2031, but the real thesis is elsewhere — software accounts for less than 0.5% of world GDP, meaning the other 99.5% of the economy is where AI's next chapter will be written.
Where does the $7.6 trillion go?
Goldman estimates global AI infrastructure investment will total roughly $7.6 trillion between 2026 and 2031, spanning compute, data centers, and power.
This means → AI spending is shifting from "train bigger models" to "build the physical base that plugs models into the real world." Compute is just the starting point; power and physical space are the larger line items.
Put simply = the money used to build the brain. Now it builds the body.
Why is "the other 99.5%" the real battleground?
Goldman notes that software represents less than 0.5% of global GDP. The remaining 99.5% — manufacturing, mining, energy, utilities — is the space AI has barely touched.
Mark Sorrell, Goldman's global head of industrials, says conversations with manufacturing executives have moved from "will the factory adopt AI?" to "how fast will automation roll out?"
This means → the industrial sector has passed the wait-and-see stage and entered execution mode. Over the next decade, many production lines are expected to rely heavily on robots, especially for hazardous tasks.
What is happening to the line between tech and industrial companies?
Jung Min, Goldman's co-head of global TMT, says AI is pulling tech firms and non-tech enterprises closer together, blurring the traditional boundary.
The report argues that what determines which industries thrive and which stall is "a capital-structure challenge more than an engineering challenge."
In plain terms = technology itself is no longer the biggest bottleneck. Whoever can organize capital faster and deploy it to the right places wins the race.
What signal is the M&A data sending?
Citing Dealogic data, Goldman notes that 2026 tech M&A volume has already reached $566 billion, surpassing the $334 billion recorded for all of 2025.
Sorrell says geopolitical tensions have not slowed deal flow, and falling global energy prices have provided additional support.
This reflects capital voting with its feet: despite a complex macro backdrop, money is accelerating into AI-linked physical assets.
What does the boardroom urgency mean?
Sorrell describes the mood inside corporate boards: "There is definitely a feeling of — 'if I don't move, will I get left behind?'"
This means → industrial AI decisions are compressing from "justification phase" to "land-grab phase." Competitive fear is replacing careful evaluation as the primary driver.
That makes the speed at which capital reaches the physical economy the key variable for whether this AI expansion actually delivers — how fast the money arrives matters more than how fast the models improve.
Content is for reference only, not financial advice.