# data center revenue — X 热门讨论 (2026-09-27 14:09 UTC)
## @XivTroy (Okong' Okuna) · 09-27 13:03 · ♥78 ↻19 💬32 History will vindicate William Ruto. The problem with metrics like economic development & political sanity is that they can only be validated in hindsight:
1. Lamu Dangote refinery = 60,000+ jobs + 500 megawatts to the national grid (feeding directly into the Olkalou Microsoft data centre, paused because of power constraints).
2. Olkalou Microsoft data center = 1000s more jobs & attendant CSR rewards for the locals.
3. 1.01 million tonnes of sugarcane supplied in July, a 102% increase from last year. Meaning more money for farmers in Western Kenya. Between Jan-May sugarcane production was at +21.98%. We'll probaby double this if the trend continue.
4. By August, Kenya was the leading milk producer in Africa with 5.5 billion litres produced. Counties like Muranga got 10,000-litre milk coolers. Farmers selling milk for 60-70, from the paltry 40 offered by the cooperatives.
5. Meat exports increased 87% in August. With almost zero red meat imports because of govt non-tariff barriers. Barriers so tight it had American exporters begging Trump to intervene. Feeding into +BOT, partly explains the stable exchange rate.
6. We have 5 concurrent major road projects. Rironi-Mau alone created 10K+ jobs. We have Bura-Mto Mwagodi in Taita-Taveta, Kakamega-Musikoma Road, Kivandini-Museleni in Machakos etc. Jobs. Direct & indirect.
7. Not to mention LAPSSET corridor, an estimated 2-3% GDP injection. Basically, +revenue for Kenya. Opening Northern Frontier. 1000s of jobs in logistics etc.
8. Turkana oil rig, 20,000+ barrels of crude oil per day if it kicks off. Estimated commencement date December 2026.
9. SHA, for all its frailties, is actively helping people. In fact, it's almost ironic, that the regions with the utmost criticism for SHA are also the counties leading in its uptake. Nairobi, Kiambu, Nakuru, Mombasa etc. The sentiments & actions don't tally.
He has his faults like everyone else, but man puts in the work.
This is the devolution you all speak of. That everyone in the country gets something. This is what economic growth looks like. And devolution, as envisaged in THE CONSTITUTION. For the champions of rule of law. https://x.com/XivTroy/status/2104195286906126466
## @jiwanpark1014 (Jay) · 09-27 05:54 · ♥33 ↻2 💬1 $NBIS Nebius and the UAE: Following the Breadcrumbs Toward a Sovereign AI Partnership
Nebius's relationship with the United Arab Emirates is beginning to look less like an isolated sales initiative and more like the early construction of a regional AI-infrastructure strategy.
There is still no publicly announced multibillion-dollar UAE government contract, no confirmed Nebius hyperscale data center in Abu Dhabi, and no disclosed partnership between Nebius and G42/Core42. Those distinctions matter.
But when the individual developments are placed chronologically, a surprisingly coherent picture emerges:
Middle East expansion → UAE commercial hiring → MBZUAI inference partnership → sovereign-AI positioning → Palantir partnership → direct engagement with the UAE's AI minister.
None of those events independently proves that a major UAE deal is coming. Together, however, they show that Nebius is deliberately positioning itself inside one of the world's most aggressive sovereign-AI ecosystems.
The first breadcrumb: Nebius explicitly targeted the Middle East
This did not begin with Roman Chernin's recent meeting.
At a Morgan Stanley conference on March 4, 2026, Nebius CEO Arkady Volozh was discussing the company's global infrastructure and commercial expansion when he said Nebius was "opening our offices in Middle East," alongside expansion into Singapore, Korea and Japan.
That comment matters because it predates many of the recent UAE developments.
Nebius was therefore already building a Middle Eastern go-to-market strategy months before the meeting with UAE AI Minister Omar Sultan Al Olama.
The company subsequently began recruiting directly for the Gulf. A recent Principal Solutions Architect–Middle East position prefers candidates based in the UAE and specifically calls for relationships with government, enterprise, AI-native and digital-native organizations throughout the Gulf. The job description also lists experience with sovereign cloud, government and regulated-industry customers as particularly relevant.
Nebius has additionally recruited a Key Customers Solutions Architect for MEA from Abu Dhabi, focused on strategic clients operating GPU workloads involving hundreds or thousands of accelerators. Earlier UAE recruitment included business development and regional field marketing roles designed to create customer pipeline across the Middle East and Africa.
This is stronger evidence than simply sponsoring a conference.
Nebius is assembling the people necessary to sell, design and support large AI systems locally.
Then came MBZUAI
The next connection is particularly interesting.
On September 3, Mohamed bin Zayed University of Artificial Intelligence's Institute of Foundation Models launched K2 Horizon, a family of six open models ranging from 0.9 billion to 375 billion parameters.
MBZUAI said K2 Horizon's APIs would be available through a group of inference partners that included AWS, Cerebras, Compass and Nebius.
That creates a direct, verifiable technical relationship between a major UAE AI institution and Nebius.
This should not be overstated: it does not mean MBZUAI has selected Nebius as its exclusive cloud provider, nor does it establish that UAE government workloads are running on Nebius.
But it does mean developers wanting to use an Abu Dhabi-developed frontier open-model family can access it through Nebius's inference ecosystem.
The timing is notable because K2 Horizon is designed specifically around reasoning, tool use and agentic workloads. MBZUAI also happens to be one of the institutions supporting the UAE government's separate program to build agentic-AI expertise across 80,000 federal employees.
Those two programs should not be conflated. There is no evidence that K2 Horizon automatically becomes the model used across UAE federal agencies.
The strategic connection is instead broader:
The same UAE ecosystem is developing open models, training government workers to build and deploy agents, and simultaneously creating relationships with inference providers such as Nebius.
That is precisely the type of environment in which an AI-infrastructure provider wants to establish itself early.
Then Roman Chernin met the UAE AI minister
The strongest relationship signal came afterward.
Nebius recently disclosed that co-founder and Chief Business Officer Roman Chernin, accompanied by Nebius's local UAE team, met H.E. Omar Sultan Al Olama, UAE Minister of State for Artificial Intelligence.
According to Nebius, the discussion concerned accelerating meaningful AI adoption and supporting Nebius's continued growth and operations in the region. Nebius also said it wanted to contribute its technology, expertise and infrastructure to the UAE's AI ambitions across businesses, governments and communities.
The language is worth examining carefully.
Nebius did not merely say Chernin attended a technology event with the minister.
The company specifically referenced:
a local UAE team, existing operations, continued regional growth, AI infrastructure, and government applications.
That makes the encounter more relevant from a due-diligence perspective.
It still is not a contract.
Government ministers routinely meet technology companies considering expansion, investment and partnerships. The meeting should therefore be treated as evidence of relationship building and strategic access, not as hidden confirmation of an undisclosed commercial agreement.
But it fits neatly into the sequence that began months earlier.
Why the UAE is strategically attractive to Nebius
The UAE is not simply buying a few AI applications.
The federal government formally established a target of moving 50% of government sectors, services and operations toward agentic-AI models within two years, including systems capable of autonomous execution and decision-making.
Implementation is already moving beyond announcements. More than 300 participants from 50 federal entities joined an implementation workshop in June, while a train-the-trainer program launched with 1,000 participants drawn from 50 ministries and government bodies.
That creates a structural compute requirement.
A conventional government digitization project might move forms from paper to websites.
An agentic-government architecture potentially requires continuous:
model inference, retrieval, database access, reasoning, tool execution, orchestration, security, logging, monitoring and human-in-the-loop workflows.
That is much closer to the infrastructure Nebius is building.
Nebius itself has increasingly moved from positioning around raw GPU capacity toward the production-agent stack. It acquired Tavily to integrate agentic web search into Nebius AI Cloud and described the move as a transition from simply running models toward running real-world AI systems.
It has also published an open Agents Blueprint for operating production AI agents and introduced Nebius Echo, an agent embedded directly into its own cloud console.
The strategic overlap is therefore straightforward:
The UAE wants AI agents deployed throughout institutions. Nebius increasingly wants to provide the infrastructure and software primitives required to operate AI agents at production scale.
The Palantir connection makes this considerably more interesting
Five days after the K2 Horizon launch, another major piece fell into place.
On September 8, Palantir named Nebius its preferred sovereign AI infrastructure partner.
Following integration, Nebius compute and inference endpoints are intended to operate inside the Palantir enterprise perimeter, allowing eligible Palantir customers greater control over their compute, data and models.
This is highly relevant to governments and regulated industries.
The most sensitive workloads cannot always be treated like ordinary SaaS applications. Governments may require particular jurisdictional controls, data residency, authorization, auditability and operational control.
Palantir provides an enterprise/government software and data-control layer.
Nebius provides the underlying compute, cloud and inference infrastructure.
Conceptually, that creates:
sovereign data + sovereign models + sovereign compute + enterprise authorization + agentic applications.
It does not establish that Palantir and Nebius are jointly pursuing the UAE government.
But it materially strengthens Nebius's product-market fit for the type of sovereign AI environment the UAE is constructing.
Nebius has also created a business model almost tailor-made for national AI programs
Another development deserves much more attention than it has received.
In July, Nebius announced a new infrastructure-partnership model specifically designed for data-center developers, infrastructure investors, regional operators and national AI projects worldwide.
The structure is different from Nebius financing every facility itself.
Under the model, a partner can finance and own the infrastructure and hardware and operate the data center.
Nebius then contributes its system architecture, supply-chain access, hardware design, cloud software and services stack and global customer distribution.
That architecture could be particularly relevant for sovereign states.
A government or domestic infrastructure investor may want local ownership of strategic AI infrastructure while still importing advanced software, hardware integration expertise and operating capabilities.
The model could therefore look like:
Local sovereign capital and infrastructure
→ locally controlled data center
→ Nebius architecture and AI-cloud stack
→ NVIDIA/other accelerators
→ domestic models and enterprise workloads
→ government and private-sector agents
This is only a possible application; no such Nebius-UAE transaction has been announced.
But Nebius itself explicitly identifies national AI projects as a target audience for the structure, making the UAE a logically relevant market.
The biggest reason not to overstate the thesis: G42 and Core42 are already deeply embedded
This is the critical counterargument.
Nebius is entering an ecosystem with powerful incumbents rather than an empty market.
Abu Dhabi already has a major sovereign-cloud partnership involving its government, Microsoft and Core42, a G42 company. That platform is intended to support more than 11 million daily digital interactions, while Abu Dhabi has stated an ambition to become a fully AI-native government by 2027.
Core42 and e& UAE have also launched an in-country Sovereign AI Compute platform for enterprise and government customers.
At the infrastructure level, G42 is leading development of a planned 5 GW UAE-US AI Campus, while Stargate UAE includes a planned 1 GW cluster, with an initial 200 MW originally expected to become operational in 2026.
So the simplistic thesis—
“The UAE needs sovereign AI, therefore Nebius will become its cloud provider”
—is not defensible.
The UAE already has Microsoft, G42, Core42, OpenAI, Oracle, NVIDIA, Cisco and others embedded throughout its AI strategy.
A more realistic thesis is that Nebius could become an additional infrastructure, inference or open-model partner inside a multi-provider ecosystem.
That would still be economically significant.
Countries building national AI capacity are unlikely to depend permanently on one model, one accelerator or one infrastructure vendor. Redundancy, model choice, performance, cost and sovereignty all encourage heterogeneous architectures.
Nebius does not necessarily need to replace Core42.
It may only need to capture a portion of incremental workloads.
Why open models matter here
Nebius's strategy increasingly emphasizes an open and independent AI stack.
Its September AI Builder Program argues that modern AI systems should combine open models, inference, orchestration, retrieval, evaluation, observability and post-training rather than forcing customers into one closed-model ecosystem.
That philosophy overlaps unusually well with K2 Horizon.
MBZUAI released K2 Horizon under Apache 2.0 and emphasized openness across weights, code and training methodology.
For sovereign customers, open models can offer several advantages: the ability to host models locally, fine-tune them on proprietary data, inspect architectures, optimize inference and reduce dependence on a single foreign API provider.
This does not eliminate security or governance issues.
But it creates a natural fit between the UAE's ambition to build its own AI capabilities and Nebius's attempt to become infrastructure for customers that want greater ownership of their AI systems.
The investment interpretation
From an NBIS perspective, the UAE story should currently be classified as strategic optionality rather than booked revenue.
The evidence for genuine regional expansion is now strong.
Nebius publicly announced Middle East office expansion months ago. It has built UAE and Gulf hiring across sales, marketing and technical architecture. It became an inference partner for MBZUAI's K2 Horizon. It created an infrastructure model aimed explicitly at national AI projects. It became Palantir's preferred sovereign-AI infrastructure provider. And its Chief Business Officer has now met the UAE's AI minister alongside a local UAE team to discuss the company's continued regional expansion.
That is a meaningful chain of evidence.
What has not yet happened publicly is equally important.
I found no announcement of a Nebius UAE hyperscale data center, no disclosed federal-government cloud award, no G42/Core42-Nebius partnership and no major UAE capacity agreement.
Those would move the thesis from ecosystem positioning toward material commercial validation.
What would validate the thesis next?
The most important future signals would be a local UAE data-center or GPU-capacity announcement; an infrastructure partnership funded by a UAE sovereign investor or domestic operator; a partnership with G42, Core42, MGX, e&, Khazna or another major local AI entity; deployment of Nebius infrastructure for UAE public-sector workloads; large Gulf enterprise customers; expansion of the MBZUAI relationship beyond model inference; or Palantir-Nebius sovereign deployments appearing in the Middle East.
Any one of those would significantly strengthen the argument.
Several occurring together would suggest that Nebius is becoming an actual infrastructure layer inside the Gulf AI buildout.
The broader picture
The most interesting part of this story is that it connects several trends that are usually discussed separately.
The UAE is attempting to turn government processes into agentic systems.
MBZUAI is developing sophisticated open models optimized for reasoning and agents.
National policy increasingly emphasizes sovereign control over compute and data.
G42 and its partners are building gigawatt-scale AI infrastructure.
Nebius is simultaneously expanding geographically, building agent infrastructure, operating open-model inference, developing sovereign-cloud partnerships and introducing a model that allows third parties to own the physical data center while Nebius supplies the technology stack.
The pieces fit.
The eventual architecture could resemble:
national power and data-center infrastructure
→ NVIDIA and other accelerators
→ Nebius AI cloud/inference layer
→ open sovereign models such as K2
→ Palantir or other enterprise orchestration layers
→ government and enterprise agents
→ citizens, workers and businesses.
This remains a thesis rather than an announced UAE architecture.
But the strategic logic is increasingly visible.
Conclusion
Roman Chernin's meeting with Omar Sultan Al Olama should not be interpreted in isolation.
By itself, a ministerial meeting tells investors little.
Placed in context, it becomes considerably more interesting.
Nebius announced Middle East expansion in March. It subsequently built a UAE-focused commercial and technical organization. MBZUAI selected Nebius as an inference partner for K2 Horizon. Nebius developed a national-AI infrastructure partnership model. Palantir selected it as its preferred sovereign AI infrastructure partner. And now Nebius's Chief Business Officer has met the UAE's AI minister with the company's local team to discuss further regional growth and AI adoption.
The appropriate conclusion is therefore not:
“Nebius has won the UAE.”
It is:
“Nebius is systematically positioning itself to participate in the UAE and wider Gulf sovereign-AI buildout, and there are now multiple independent breadcrumbs showing that this strategy has moved beyond theory.”
The next stage is commercial proof.
If Nebius converts these relationships into local capacity, sovereign infrastructure partnerships or major government and enterprise workloads, the UAE could evolve from a small geographic expansion story into an important part of Nebius's ambition to become a global AI-native hyperscaler.
Until then, it is one of the more interesting pieces of optionality inside the NBIS thesis—and one worth watching closely. https://x.com/jiwanpark1014/status/2104087217606209566
## @NickHorob (Nick Horob) · 09-27 03:31 · ♥31 ↻0 💬1 This is a 5-well pad drilled right on the North Dakota-Montana border on one of my cousin's farms (no mineral acres of mine).
Let's calculate the operating profit if you install a "small" data center on site that uses the natural gas to power it.
This well pad has produced 50,000 MCF of natural gas on average in 2026. Very little decline rate.
This could support an 8- to 9 MW natural gas generator, but I'm going to be conservative and size it down to 5 MW.
Capex: - NG generator: $15 million - Data center and infrastructure: $75 million - GPUs (~2,500 Nvidia B200): $120 million Total = $210 million
Revenue @ 50% token utilization ($15M output/ $3M input) = $190 million/yr
Cash Expenses = ~$15 million/yr
Depreciation: - GPUs at 6 years = $20 million/yr - Data center @ 15 years = $6 million/yr
Operating profit = $149 million/yr (71% annual ROI)
The problem is that there's an insane backlog for things like natural gas generators. And there's a ton of uncertainty about AI tokenomics. https://x.com/NickHorob/status/2104051210441679310