CoreAI
AI cloud infrastructure
AI / MLB2BTashkent
CoreAI is an AI infrastructure platform that helps companies in Central Asia deploy, train and serve generative AI and ML models in production. One platform offers rentable GPU servers, a ready-to-use AI chat, pay-as-you-go access to AI models and custom training on a company's own data, with no in-house technical team needed.
Traction
$12,700
MRR, Sep 2026
$58,400
Revenue to date
3
Clients + 2 pilots
$160K+
Own GPU fleet
- Monthly revenue up 4.2x since February, from $3,000 to $12,700
- Paying clients include Beeline, Aiphoria and the Agro-industry Digitalization Center
- GPU fleet now worth $160K+ as market prices rose; H200-class hardware in production
The problem
Closed model APIs are easy to start with but costly at scale, offer no control and send data abroad. Self-hosting open models needs GPUs and rare skills, and the region has no platform for it.
- Closed model APIs get expensive at scale and send data abroad
- Self-hosting open models needs GPUs and rare engineering skills
- Projects burn money on APIs or stall on infrastructure
- No regional platform to rent GPUs, deploy models and fine-tune in one place
The solution
Full-stack AI cloud: GPU rental plus managed inference, fine-tuning and data collection, so clients reach production without an AI team. H200-class hardware already runs in production.
- GPU rental - On a monthly contract
- Inference API - Pay-per-token, first in the region
- Fine-tuning - On the client's own data
- Data collection - As an enterprise service
The company
What makes it different
Working AI, not raw GPUs
Regional players moved from traditional cloud to GPU rental. CoreAI adds managed inference, fine-tuning and data collection on top, so clients reach production without building an AI team, with local accountability.
Who it's for
- Primary: AI builders in Central Asia now using OpenAI or renting foreign GPUs
- Secondary: enterprises with data-residency constraints, already a revenue-generating segment
How it makes money
- GPU rental: monthly contract, A100 80GB from about UZS 24M a month
- Inference API: self-hosted, usage-based
- Enterprise services: fine-tuning and data collection, the highest-margin line
Market
$673M
Uzbekistan cloud market today
$2.08B
Uzbekistan cloud, 2030 (25% CAGR)
$26.6B
Global GPU-as-a-service
$255B
AI inference market by 2030
Open models such as Qwen, Gemma and DeepSeek have caught up with closed APIs on quality, but the region has no local AI cloud, pay-per-token API or managed fine-tuning. Demand has arrived: enterprises are running their first AI infrastructure tenders.
Competitors
- Foreign - AWS, GCP, Azure and GPU clouds such as RunPod, Vast.ai, Together AI: no local accountability or data residency.
- Local - UzCloud, Pro Data-Tech, IT Global: moved from traditional cloud to GPU rental, with no AI stack layer.
Founders
Sanjarbek has worked with AI and LLMs since 2019 and has fine-tuned many LLMs that serve real users. At CoreAI, Sanjarbek handles all AI technical work: inference optimisation, fine-tuning, model evaluation and benchmarking.
After a full-scholarship Computer Science degree at KAIST, spent about two years at an AI startup, rising to Lead AI Engineer over a team of four. Then an AI Master's at FAU Erlangen-Nürnberg, alongside work at Alstin Capital, a €175M VC fund.
Left the Master's early to work on CoreAI full time. The team has brought GPUs through customs, built the serving stack and won paying enterprise customers through competitive tenders; H200-class hardware runs in production.
Sherzod Safarov
CTO at CoreAI
Background
- KAIST, South Korea: Computer Science, Bachelor's, full scholarship
- FAU Erlangen-Nürnberg, Germany: Master's in AI, left early to build CoreAI
- Lead AI Engineer at an AI startup, ~2 years, leading four engineers
- Alstin Capital, Germany: a €175M early-stage venture fund
- Working with AI and LLMs since 2019
Milestones
- 3 SepBeelab trialled the service; 30 people registered for chat and inference; Private Cloud proposal sent to a bank
- 10 SepContract signed with Beelab; investor due diligence under way
- 14 SepNamed reserve supplier in Beeline's managed-inference tender, a project 10x larger than its previous tender win
- 15 SepPitched at IT Park for the Presidential AI Award
- 21 Sep3 paying customers plus 2 pilots; $48,000 in revenue to date; MRR $12,700
- Oct$58,400 in revenue to date; Beeline among the paying clients; GPU fleet worth $160K+
The round
$200,000
raising
Use of funds
GPU fleet expansion
55%
Platform, AI and Confidential Cloud launch
32.5%
Team and enterprise sales
12.5%
