MBZUAI: Inside the World's First AI University in the UAE

What Is MBZUAI?

The Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) is a research university established in 2019 in Masdar City, Abu Dhabi, and it describes itself as the world's first graduate-level, research-based AI university. It launched with master's and PhD programs only, appointed Carnegie Mellon professor Eric Xing as founding president in November 2020, and now anchors UAE AI research.

Key Takeaways

• MBZUAI, founded in 2019 in Masdar City, Abu Dhabi, calls itself the world's first graduate-level, research-based AI university.
• The university states that, as of 2024, it ranks in the global top 10 in core AI fields on CSRankings.
• Its Fall 2025 intake enrolled 403 students from more than 8,000 applications, a roughly 5 percent acceptance rate, per figures reported by Zawya.
• With G42 and Cerebras, MBZUAI released the K2 Think reasoning system in September 2025 and K2 Think V2, a 70B sovereign model, in January 2026.
• MIT's Schwarzman College of Computing and MBZUAI launched a five-year joint research program in October 2025.

Most universities added AI the way homeowners add an extension: a new wing bolted onto a structure designed for something else. MBZUAI poured its foundation around AI from day one. Every department, hiring decision, and square meter of lab space serves a single field. That is the institutional bet Abu Dhabi made in 2019, and it is central to the story we cover in our complete guide to AI in the UAE.

The leadership choice signaled ambition early. Eric Xing, a machine learning heavyweight from Carnegie Mellon, took the presidency in November 2020 and still holds it as of mid-2026. Under his tenure the university has recruited faculty aggressively and, by its own account, climbed the rankings fast.

As of 2024, MBZUAI states it has entered the global top 10 in core AI fields on CSRankings, a claim worth noting precisely because the institution is barely five years old. CSRankings scores universities on publication output at selective venues, so the figure reflects research volume and placement rather than reputation surveys. Treat it as self-reported, but treat it seriously.

What Does MBZUAI Teach?

MBZUAI offers master's and PhD programs across machine learning, computer vision, natural language processing, robotics, computer science, computational biology, and statistics and data science, organized into eight research departments, according to the university. In March 2025 it unveiled its first undergraduate degree, a BSc in AI, with the first class starting Fall 2025.

The graduate-only origin was deliberate. A new institution cannot compete with century-old universities on breadth, so MBZUAI competed on depth: small cohorts, funded research degrees, and faculty hired for one field. For six years, that meant no undergraduates at all.

That changed in 2025. The BSc in Artificial Intelligence arrived with two streams, Business and Engineering, a structure that says a lot about the intended output. Abu Dhabi does not just want model builders. It wants people who can carry AI into boardrooms and government departments, a theme that runs through how the UAE teaches AI at every level of its education system.

Demand has been striking. The Fall 2025 intake enrolled 403 new students, the university's largest cohort and its first with undergraduates, from more than 8,000 applications, an acceptance rate of roughly 5 percent, according to MBZUAI figures reported by Zawya. For context, that selectivity band overlaps with the most competitive universities anywhere. Whether it holds as the undergraduate program scales is a question worth revisiting in a year.

What Does MBZUAI Research?

MBZUAI's research runs through dedicated institutes: the Institute of Foundation Models, the Institute of Digital Public Health, and the Institute for Agriculture and AI, with a Robotics Research Lab opening in 2026, per the university. Its most visible output is the K2 Think reasoning series, built with G42 and Cerebras, whose 70B second version shipped in January 2026.

The institute structure matters because it converts academic output into products with names. Papers are the currency of rankings; released models are the currency of relevance. MBZUAI has decided it wants both.

K2 Think: a reasoning model with a sovereignty pitch

On September 9, 2025, the Institute of Foundation Models and G42 launched K2 Think, a 32B open-source reasoning system that the partners said matches flagship reasoning models roughly 20 times its size, running on Cerebras hardware. The efficiency claim was the headline: small model, big reasoning.

Version two raised the stakes. On January 27, 2026, MBZUAI, G42 and Cerebras released K2 Think V2, a 70B system described as fully sovereign and built on the K2-V2 base. The scores are specific: 90.42 on AIME 2025 and 72.98 on GPQA Diamond, per MBZUAI.

Why does "fully sovereign" matter here? Because it means the base model, not just the fine-tuning, was produced within the UAE ecosystem. For a government that treats AI capability as strategic infrastructure, owning the full stack is the point. An investor reading a term sheet for a Gulf AI startup in 2026 will notice that the region now trains frontier-adjacent models at home rather than renting them.

MBZUAI is not alone in this. The Technology Innovation Institute (TII) runs a parallel open-model program, and its Falcon-H1R 7B reasoning model, released January 5, 2026, scored 88.1 percent on AIME-24 per TII. We profile that lineage separately in our piece on Falcon models and TII. Two publicly backed Abu Dhabi labs shipping competitive reasoning models within the same month is not a coincidence; it is a strategy.

What Is Jais, and Where Does MBZUAI Fit?

Jais is an open Arabic large language model announced on August 30, 2023, billed at release as the world's most advanced open Arabic LLM. It was built by Inception, a G42 company, in collaboration with MBZUAI and Cerebras, trained on the Condor Galaxy supercomputer with 116 billion Arabic and 279 billion English tokens.

MBZUAI's role was academic muscle inside a commercial build. The original Jais release paired G42's compute and productization with university research talent, an early template for the K2 Think collaboration that followed two years later.

The Arabic data problem explains why this mattered. High-quality Arabic text is scarce online relative to English, so the 13B Jais model's 116 billion Arabic tokens represented serious curation work, not just scraping. Jais demonstrated that a UAE academic-commercial partnership could set the global benchmark for an entire language's AI tooling.

G42 then scaled the family. On August 5, 2024, it launched Jais 70B alongside a suite of 20 models ranging from 590M to 70B parameters, trained on 370 billion tokens including 330 billion Arabic, the largest Arabic dataset used for an open foundation model at the time, according to G42. MBZUAI's fingerprints on the original architecture remain part of that story, even as the commercial line moved to G42.

How Does MBZUAI Connect Globally?

MBZUAI's flagship international tie is with MIT. In October 2025, MIT's Schwarzman College of Computing and MBZUAI launched a five-year collaborative research program focused on scientific discovery, human thriving, and planetary health, led by MIT's Philip Isola and MBZUAI's Le Song, according to MIT News.

The MIT partnership does two things for Abu Dhabi. It validates the young university in front of the global academic market, and it gives MBZUAI faculty and students a standing channel into one of the field's densest research networks. Five years is long enough for joint PhD pipelines and repeated publication cycles, not just a press release.

There is a recruiting angle too. A researcher weighing an offer in Masdar City can now point to formal MIT collaboration when colleagues ask about isolation from the field's centers of gravity. In our experience tracking Gulf institutions, that kind of signal moves hiring conversations more than salary premiums do.

International faculty recruitment was already central to the model. The founding president came from Carnegie Mellon, and the CSRankings claim rests on publication patterns typical of researchers recruited from top Western and Asian programs. The MIT program formalizes what the hiring strategy implied.

The Wider UAE Research Bench

MBZUAI is the most visible node, but UAE AI research runs through several universities. Khalifa University's TelecomGPT-R1, an open 27B reasoning model for telecom released under Apache 2.0, topped the GSMA Open Telco Leaderboard with an 89.6 percent average after launching at MWC Shanghai in June 2026, according to Khalifa University. UAEU in Al Ain teaches a BSc in Data Science and AI.

Khalifa University's result deserves attention because it is a vertical play. Rather than chasing general benchmarks against far larger labs, its team built a domain reasoning model for telecommunications and, per the university's reported figures, beat the field in that domain as of June 2026. Specialized open models are a lane where mid-sized research universities can win outright.

United Arab Emirates University covers the feeder layer. Its College of IT offers a BSc in Data Science and Artificial Intelligence plus an AI minor, according to the UAEU catalog, giving Al Ain students a domestic on-ramp that did not require Abu Dhabi or an overseas degree.

The pattern across the three institutions is complementary rather than duplicative. MBZUAI aims at frontier research and graduate depth, Khalifa University at engineering and domain models, UAEU at broad undergraduate supply. Add TII's Falcon program on the applied-research side and the emirates field a deeper bench than any single headline suggests.

Why Does a Dedicated AI University Matter for the Talent Pipeline?

A dedicated AI university converts national strategy into people. MBZUAI's reported 403-student Fall 2025 intake, drawn from over 8,000 applications per Zawya's reporting, feeds a labor market where AI skills carry substantial salary premiums, as we detail in our guide to AI jobs and salaries in the UAE.

Talent is the binding constraint on every AI strategy, and the UAE has historically imported it. A domestic institution producing PhD-level researchers changes the arithmetic. Graduates who studied in Masdar City have local networks, local research relationships, and in many cases funded ties to local employers such as G42 before they finish.

The K2 Think projects show the loop closing. Students and faculty at the Institute of Foundation Models work on models that G42 productizes and the government can deploy, so training, research, and industry absorption happen inside one small geography. The university is less a school than a talent refinery placed deliberately at the center of the national AI supply chain.

None of this guarantees retention. Top graduates receive global offers, and the UAE competes for its own alumni like everyone else. But the direction is clear: each cohort of a few hundred highly selected students compounds, and the first undergraduates who entered in Fall 2025 will hit the labor market just as the current buildout of compute and applications matures.

Frequently Asked Questions

What does MBZUAI stand for?

MBZUAI stands for the Mohamed bin Zayed University of Artificial Intelligence. Established in 2019 in Masdar City, Abu Dhabi, it describes itself as the world's first graduate-level, research-based AI university. It began with master's and PhD programs only, and added its first undergraduate degree, a BSc in AI, with the first class starting in Fall 2025.

Is MBZUAI really the world's first AI university?

MBZUAI is the first university built from scratch as a graduate-level, research-based institution dedicated entirely to artificial intelligence, according to the university itself. Other universities ran strong AI departments long before 2019, but none had organized an entire degree-granting institution around the field. The claim refers to that institutional focus, not to who researched AI first.

How hard is it to get into MBZUAI?

Very selective, based on reported figures. MBZUAI's Fall 2025 intake enrolled 403 new students, its largest cohort and the first to include undergraduates, drawn from more than 8,000 applications, an acceptance rate of roughly 5 percent, according to university figures reported by Zawya. That places it in the same selectivity band as elite US institutions.

What is K2 Think?

K2 Think is an open-source AI reasoning system from MBZUAI's Institute of Foundation Models and G42, launched September 9, 2025 as a 32B model said to match flagship reasoning models roughly 20 times its size, running on Cerebras hardware. K2 Think V2 followed on January 27, 2026: a 70B fully sovereign system scoring 90.42 on AIME 2025.

Who leads MBZUAI?

Eric Xing, a Carnegie Mellon University professor, was appointed founding president in November 2020 and remains president as of mid-2026. Under his leadership the university states it has entered the global top 10 in core AI fields on CSRankings as of 2024, and it launched research institutes covering foundation models, digital public health, and agriculture.

How does MBZUAI relate to the Jais Arabic language models?

Jais, announced on August 30, 2023 and billed at release as the world's most advanced open Arabic LLM, was built by Inception, a G42 company, in collaboration with MBZUAI and Cerebras. The 13B model trained on 116 billion Arabic and 279 billion English tokens. G42 expanded the family in August 2024 with Jais 70B and a suite of 20 models.

Where to Go Next

MBZUAI is one pillar of a larger structure. For the full picture, start with our complete guide to AI in the UAE, then follow the applied-research thread through Falcon models and TII and the education thread through how the UAE teaches AI. If you want new briefings like this one as they publish, subscribe via our contact page.