GS has raised its 2030 global data center capacity forecast from 217 GW to 273 GW a massive 26% upgrade.
That’s an estimated $8.5 TRILLION in infrastructure spending over five years.
And here’s the real story: AI demand isn’t slowing down. The race for power, GPUs, and data center capacity is accelerating.
The biggest AI investment cycle may still be ahead of us. ๐งต๐
— Exencial Research Partners (@exencial_rp) Nov 03 04:57
An X account posted this on 3 November with an image of four Goldman Sachs Global Investment Research exhibits. The post says Goldman raised its 2030 forecast from 217 GW to 273 GW and cites an estimated $8.5 trillion of spending over five years. We used the exhibits’ printed figures, checked the 217 GW against July coverage and the $8.5 trillion against CNBC; the post’s views on demand are its own.
Goldman Sachs now puts global data center capacity at about 273 GW by 2030, according to exhibits from a Goldman Sachs Global Investment Research report posted on X on 3 November. In July, Investing.com reported a Goldman forecast of 217 GW for 2030. The new figure is 56 GW, or about 26%, higher, by our arithmetic.
Goldman also put the bill at about $8.5 trillion in a report on Monday, CNBC reported; the material we read does not show whether that is the same report as the exhibits. “Global data center demand will require about $8.5 trillion in capital spending over the next five years, Goldman Sachs wrote in a report Monday”, CNBC reported on 3 November. Goldman’s own yearly figures for new capacity add up to about 170 GW from 2026 to 2030, or about 34 GW a year, by our arithmetic.
The key insight: As we read it, Goldman’s 273 GW is a build rate as much as a total: its yearly net adds average about 34 GW from 2026 to 2030, by our arithmetic, and rise every year to 40.0 GW in 2030, while consensus capex growth for eight AI and cloud spenders falls to 2% in the same image.
What the Exhibits Show
Exhibit 1 is titled “Global Data Center Capacity should scale to ~273 GW by 2030 (+170 GW v. 2025)”. It measures capacity in megawatts and cites 451 Research and Goldman Sachs Global Investment Research as sources. Subtracting the 170 GW from 273 GW puts the 2025 base at about 103 GW, by our arithmetic; the July coverage gave 101 GW for 2025.
Exhibit 2 shows how often the forecast has moved. Its title says the 2026-2030 forecasts are “+11% on average v. 1Q26, +17% on average v. 4Q25, and +26% v. 3Q25”. The July report, Investing.com wrote, had already raised the 2030 figure from 168 GW to 217 GW, citing project activity tracked by 451 Research.
Exhibit 3 gives the yearly net adds. They rise from 18,603 MW in 2025 to 25,117 MW in 2026, 32,466 MW in 2027, 34,797 MW in 2028, 37,295 MW in 2029 and 39,973 MW in 2030. The five estimated years sum to 169,648 MW, which matches the 170 GW in Exhibit 1, by our arithmetic.
Global data center capacity net adds per year in gigawatts (Exhibit 3 of a Goldman Sachs Global Investment Research report, sourced to 451 Research; 2026 to 2030 are estimates) and year-on-year growth in Visible Alpha consensus capex for Apple, Amazon, Meta, Alphabet, Microsoft, Oracle, SpaceX and CoreWeave (Exhibit 10, as of 30 October 2026), both from exhibits posted on X on 3 November. The 34 GW yearly average is our arithmetic.Animated: global data center capacity net adds per year, from 18.6 GW in 2025 to an estimated 40.0 GW in 2030, as printed in Exhibit 3 of a Goldman Sachs Global Investment Research report (source line: 451 Research, Goldman Sachs Global Investment Research), from an image posted on X on 3 November. The 2026-2030 adds sum to about 170 GW, about 34 GW a year (our arithmetic). Chart: FourWeekMBA.
Big Sites Take Two-Thirds of Adds From 2028
Exhibit 3 also splits each year’s adds by project size. Builds of 75 MW or more were 25% of net adds in 2025. In Goldman’s estimates they are 27% in 2026, 47% in 2027 and 66% in each year from 2028 to 2030.
Goldman’s exhibit says projects of that size are “typically assoicated with High-Performance Computing data centers and AI workflows” (the spelling is the exhibit’s). On its shares, about 96 GW of the 170 GW added from 2026 to 2030 comes from such builds, about 57%, by our arithmetic.
Adds Rise, Capex Growth Fades
The post also shows Exhibit 10, titled “AI/Cloud player spend should grow to ~$1.74 trillion by 2030 at a 31% 2025-2030 CAGR”. It lists consensus capex estimates from Visible Alpha, as of 30 October 2026, for Apple, Amazon, Meta, Alphabet, Microsoft, Oracle, SpaceX and CoreWeave.
The eight companies’ total is $918.2 billion for 2026, $1.52 trillion for 2027 and $1.74 trillion for 2030. Year-on-year growth in that total is 101% in 2026 and 65% in 2027, then 10%, 3% and 2%. Over 2026 to 2030 the consensus figures add up to about $7.55 trillion, by our arithmetic.
Set side by side, the two exhibits move differently after 2027. Goldman’s yearly net adds keep rising, from 32.5 GW in 2027 to 40.0 GW in 2030, while consensus capex growth for the eight spenders falls to single digits. The two series come from different sources and measure different things: megawatts of capacity worldwide, and the capital spending of eight companies.
What a Gigawatt Costs on Goldman’s Numbers
In July, Investing.com reported that the additional 116 GW of capacity “would require roughly $6 trillion in capital spending”. That is about $52 billion per gigawatt, by our arithmetic. Setting the $8.5 trillion over five years against the exhibits’ 170 GW of net adds gives about $50 billion per gigawatt, if both come from the same analysis and cover the same window, which the material we read does not show.
On those rough ratios, the dollar estimate rose about 42% while the gigawatts rose about 47%, by our arithmetic. The coverage we read does not say whether the $8.5 trillion includes chips and servers, power generation or only buildings.
Two Different Gigawatts
Other estimates use other measures. SemiAnalysis chief executive Dylan Patel estimated that the world adds about 40 GW of compute in 2027, as we reported, and said the world was at “over 100” GW by the end of this year. Goldman’s net adds for 2027 are 32.5 GW, about 7.5 GW lower, by our arithmetic.
The two are not like for like. Patel spoke of compute, AI and non-AI, and did not say whether he counts IT power or whole-site power. Goldman’s exhibits count data center capacity in megawatts from 451 Research data and do not define the measure in the text we read.
Who Pays
CNBC cited the Goldman figure in a report on a Hong Kong panel. There, Blackstone president Jon Gray said financing the buildout would need public markets, banks and private capital, and Pimco chief executive Emmanuel Roman cited an estimate of around $10 trillion of AI-related capital spending over five years, according to CNBC.
Business Pill · Capex and Depreciation
An evergreen Business Engineer explainer on capex and depreciation: paying once for equipment and spreading its cost over its life. It teaches the general idea only and does not discuss Goldman Sachs or data center forecasts.
As we read Exhibit 3, the raised forecast is also about site size: builds of 75 MW or more go from 25% of net adds in 2025 to 66% in each year from 2028 in Goldman’s estimates, so most of the added gigawatts sit in the large campuses its exhibit ties to AI workloads.
As we read the two exhibits together, consensus capex for the eight named spenders grows 10%, 3% and 2% from 2028 to 2030 while Goldman’s yearly adds keep rising; the exhibits do not say who funds the difference, and they measure different things.
As we read the dollar figures, the price per gigawatt barely moved on Goldman’s numbers, about $52 billion in July and about $50 billion now, by our arithmetic and assuming the same window; the larger bill reflects more gigawatts rather than a higher cost for each.
Jon Gray, Blackstone president, as reported by CNBC, 3 November 2026
“The creativity is needed because the scale of capital is enormous”
Three Implications
FOR THE BIGGEST SPENDERS Visible Alpha consensus in Exhibit 10 puts the eight companies’ capex at $918.2 billion in 2026 and $1.74 trillion in 2030, with growth of 101% in 2026 and 2% in 2030.
FOR LARGE-SITE BUILDERS Goldman’s Exhibit 3 has builds of 75 MW or more at 66% of global net adds in each year from 2028 to 2030, up from 25% in 2025.
FOR LENDERS AND INVESTORS CNBC reported Goldman’s $8.5 trillion estimate alongside Blackstone’s Jon Gray saying the buildout needs public markets, banks and private capital, and Pimco’s Emmanuel Roman citing around $10 trillion over five years.
The framework states: “The structural bottleneck isn’t computing, it’s electrons.”
As we read it, Goldman’s exhibits count the buildout in megawatts, not chips: about 170 GW of net adds from 2026 to 2030, two-thirds of them in builds of 75 MW or more from 2028, by its shares. The framework was published in December 2025 and does not discuss this Goldman report.
What Is Not Established
We did not read the Goldman report itself. The 273 GW, the net adds and the size shares come from four exhibits in an image posted on X, with Goldman and 451 Research source lines. The $8.5 trillion is as reported by CNBC, and we cannot confirm it comes from the same report. The 217 GW is as reported by Investing.com in July. The report’s title, authors, regional split and the scope of the $8.5 trillion are not in the material we read. Exhibit 10 is consensus data from Visible Alpha, not a Goldman forecast. We did not contact Goldman Sachs.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Goldman’s exhibits put global data center capacity at about 273 GW by 2030, 26% above the 217 GW reported in July, with net adds averaging about 34 GW a year from 2026, by our arithmetic. In the same image, consensus capex growth for eight AI and cloud spenders falls to 2% by 2030 while the yearly adds keep rising.
A note on sourcing. We did not read the Goldman Sachs report itself. Exhibit figures are transcribed from the printed text of an image of four Goldman Sachs Global Investment Research exhibits posted on X on 3 November; we did not read bar heights. The $8.5 trillion is as reported by CNBC and the July 217 GW forecast as reported by Investing.com. Exhibit 10 is Visible Alpha consensus data. Sums, ratios and differences are our arithmetic. We did not contact Goldman Sachs.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
