AI infrastructure investment 2026 UAE
Goldman Sachs published its most precise AI infrastructure forecast of the year in May 2026, and the headline number is worth sitting with for a moment before moving on.
$7.6 trillion. That is the projected cumulative capital expenditure on AI infrastructure between 2026 and 2031. Compute, data centres, and power. That single figure is roughly equivalent to a quarter of annual US GDP, and it is more than Germany’s entire annual economic output. It is also, Goldman says, larger than the total cumulative investment in the global internet across the entire 1990s and 2000s.
The world is about to spend more money building AI infrastructure in six years than it spent building the internet in twenty. The UAE decided five years ago that it would be a destination for that money, not just a consumer of what it creates. Here is how that bet looks today.
| VERDICTThe UAE is one of the most significant AI infrastructure destinations on earth.Stargate UAE, G42, Mubadala, and the OpenAI facility represent a real slice of the $7.6 trillion build. The UAE bet early and the bet is paying. The questions now are whether it can convert infrastructure into genuine AI capability, whether the energy and water costs of running it are accounted for honestly, and whether it remains a node in the global network or becomes dependent on it. |
What the $7.6 Trillion Actually Breaks Down To
The Goldman figure is not vague. It comes with a specific breakdown.
Compute accounts for $5.1 trillion of the total. Goldman assumes Nvidia captures 75% of all compute spend over the six-year period, roughly $3.8 trillion through a single company’s products. The baseline chip in their model is the Rubin VR200, priced at $80,500 per GPU. At 75% gross margins on data centre GPUs, Goldman describes Nvidia as collecting the largest infrastructure toll in the history of technology.
Data centres account for $2.1 trillion. The physical specifications are shifting dramatically. Standard cloud infrastructure runs at 5 to 15 kilowatts per rack. The Nvidia rack that Goldman uses as the baseline in its model runs at roughly 130 kilowatts. That is a ten-fold increase in power density per rack, which means every data centre building requires completely different engineering: industrial-scale liquid cooling, dedicated power delivery, and redundancy systems that did not exist five years ago.
Power accounts for $358 billion, just 5% of total spend but what Goldman calls the critical bottleneck for all the rest. Data centre occupancy rates are projected to hit 95% by late 2026 before new capacity begins to moderate them. Amazon alone is projected to spend $125 billion on AI infrastructure in 2026, with 3.8 gigawatts of power capacity locked in. Meta signed a 20-year, 2,600-megawatt nuclear deal with Vistra to secure the power its AI needs.
Goldman’s baseline projects $765 billion in annual AI capex in 2026, growing to $1.6 trillion annually by 2031. Annual spending doubles in six years.
Where the UAE Sits in This
The UAE is not a spectator. It is an active node in the build, and a more significant one than most coverage suggests.
The official OpenAI Stargate UAE announcement describes a one-gigawatt compute cluster in Abu Dhabi, built by G42 and operated with OpenAI, Oracle, Nvidia, and Cisco. The first 200-megawatt phase is expected online by the end of 2026. The UAE’s AI minister has described the broader Abu Dhabi site as eventually reaching five gigawatts of total capacity, larger than Monaco, at a cost of more than $30 billion. Whether measured at the first phase or the full vision, it is among the largest AI infrastructure projects on earth.
G42, the Abu Dhabi-based group chaired by the UAE National Security Adviser, is the on-the-ground builder and operator. The Microsoft investment of $1.5 billion in G42 in 2024, which came with a board seat and a condition to unwind Chinese technology partnerships, placed the UAE firmly in the US-aligned AI infrastructure network. That alignment matters because Nvidia chips, which account for $3.8 trillion of the $7.6 trillion Goldman projection, follow US export rules. Being in the US camp means access to the hardware that runs the modern AI stack.
Mubadala, Abu Dhabi’s sovereign wealth fund, has invested across the AI infrastructure chain globally, with positions in semiconductor companies, data centre operators, and AI labs. This is capital from the UAE flowing into the global AI build from the investment side as well as the hosting side.
The Honest Questions
Goldman’s own report makes clear that the $7.6 trillion figure is highly dependent on assumptions that could shift. The economic lifespan of AI chips is the largest variable. If chips depreciate over three years instead of five, annual depreciation costs rise by nearly a trillion dollars. The model is built on assumptions about Nvidia’s dominance that could be disrupted by custom silicon from hyperscalers or by advances in competing chip architectures.
And the report notes something that does not make the investment brochures. OpenAI is projected to lose $14 billion in 2026. Products like GitHub Copilot remain subsidised. The ultimate bill for the multi-trillion-dollar wager has not yet been assigned. The infrastructure is being built ahead of the revenue that is supposed to justify it. That does not make it wrong. But it means the people building the biggest AI campus in the Gulf are making a bet on future demand, not executing against guaranteed orders.
The energy question is also not going away. Goldman projects a 165% increase in global data centre power consumption by 2030. A five-gigawatt campus in Abu Dhabi, a desert city with a net-zero-by-2050 target, requires more power than most mid-sized countries use. The UAE has nuclear power at Barakah and solar at scale. It also has an honest reckoning to do about whether AI infrastructure and sustainability commitments can coexist at this build rate. The answer is probably yes, with effort. It is not automatic.
What It Means for You
If you are a UAE resident, the $7.6 trillion build affects you in ways that compound over the next five years.
The compute being built in Abu Dhabi means the UAE’s government AI services, healthcare AI, financial AI, and smart city systems are running on infrastructure that exists locally rather than depending on data centres in Virginia or Frankfurt. That matters for speed, for data residency, and for the UAE’s ability to run its own AI stack independent of foreign infrastructure owners.
It also means the UAE is in the conversation when the world’s biggest technology companies decide where to put the next phase of their AI build. Being a trusted, infrastructure-capable, US-aligned destination with stable governance, generous incentives, and access to chips is a real competitive advantage. The UAE earned that position over the last five years by deciding loudly and early that it would. The Goldman number is the proof that the decision was timed right.
The caveat is the same one this guide always carries. Infrastructure is not capability. A country can host the world’s fastest road and still not produce the cars. The UAE’s challenge in the $7.6 trillion era is to convert its position as a host into genuine domestic AI capability, which means researchers, engineers, Arabic-native models, and applications built for the region rather than adapted from elsewhere. The infrastructure is in place. The human capital build is the next phase, and it is harder.
Sources
• Goldman Sachs: Tracking Trillions, the assumptions shaping the scale of the AI build-out — https://www.goldmansachs.com/insights/articles/tracking-trillions-the-assumptions-shaping-scale-of-the-ai-build-out
• Blockonomi: Goldman Sachs maps $7.6 trillion AI infrastructure spending through 2031 — https://blockonomi.com/goldman-sachs-maps-7-6-trillion-ai-infrastructure-spending-through-2031/
• FourWeekMBA: Goldman Sachs just mapped where $7.6 trillion goes — https://fourweekmba.com/goldman-sachs-7-6-trillion-ai-capex-map-of-ai-nvidia-infrastructure/
• Goldman Sachs: AI to drive 165% increase in data centre power demand by 2030 — https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030
• UAE AI Minister Omar Sultan Al Olama: Stargate UAE five-gigawatt full site vision (January 2026 statements)
• OpenAI: Stargate UAE official announcement, one-gigawatt cluster, first 200MW phase by end 2026
Robius.news — Dubai, UAE — 2026 | Built to be first. Built to be trusted.
