Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
NVIDIA announced on October 23, 2024, that Indian infrastructure providers would add tens of thousands of NVIDIA Hopper GPUs to build large-scale “AI factories” in India. NVIDIA said the expansion could deliver nearly 180 exaflops of cumulative computing capacity and increase NVIDIA GPU deployment in the country by nearly 10 times compared with 18 months earlier. The initial infrastructure partners were Yotta Data Services, Tata Communications, E2E Networks and Netweb Technologies. A separate NVIDIA–Reliance Industries partnership targeted AI services for Jio and AI-ready data centers that could eventually reach 2,000 megawatts.
That announcement described an ecosystem of providers and customers—not one NVIDIA-owned supercomputer, a guaranteed public GPU pool or free access for every Indian business. By February 2026, NVIDIA was describing a broader program involving Blackwell systems, additional infrastructure partners and India’s sovereign-AI ambitions.
What NVIDIA actually announced
The announcement was made during NVIDIA’s AI Summit in Mumbai, held from October 23 to 25, 2024. NVIDIA said Indian infrastructure companies would deploy tens of thousands of Hopper GPUs for model training, fine-tuning and inference. The company identified four initial infrastructure leaders:
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
- Yotta Data Services
- Tata Communications
- E2E Networks
- Netweb Technologies
NVIDIA described the planned buildout as nearly 180 exaflops of cumulative computing capacity and said India’s NVIDIA GPU deployment would grow by nearly 10 times by year-end compared with the level 18 months earlier. Those are NVIDIA’s figures and projections, rather than an independently verified count of operational GPUs available to customers.
#1 Best Overall
The announcement also involved NVIDIA’s accelerated-computing platform, networking and software. The local companies would own, operate, integrate or commercialize different parts of the infrastructure, while customers could access capacity through cloud services, hosted systems or dedicated installations. NVIDIA’s original announcement is available in its India AI infrastructure overview.
What an “AI factory” means
An AI factory is NVIDIA’s industrial metaphor for a data center that turns data and computing power into AI output. It does not manufacture semiconductor chips.
A typical AI factory combines:
- GPU servers or GPU-based superchips;
- high-speed GPU-to-GPU networking;
- high-performance storage and data pipelines;
- power, cooling and data-center facilities;
- software for training, fine-tuning, inference and orchestration; and
- systems for monitoring, scheduling and serving models to users.
The distinction matters. A large GPU order by itself does not create useful AI capacity. If storage cannot feed the GPUs, networking slows distributed training, or software and scheduling are poorly configured, expensive hardware may sit idle. NVIDIA’s broader explanation of AI factories emphasizes the combination of accelerated computing, networking and software rather than chips alone.
Recommended Free Tools
Who is building the infrastructure?
| Company | Role in the announced ecosystem | Potential customer use |
|---|---|---|
| Yotta Data Services | Operates Shakti Cloud, a managed GPU-cloud platform. The 2024 plan involved thousands of Hopper GPUs and NVIDIA AI Enterprise. | Training, fine-tuning and inference for language generation, biomolecular applications and virtual avatars. |
| Tata Communications | Planned a large Hopper-GPU deployment for public-cloud infrastructure, combined with Tata’s AI Studio and network. | Enterprise workloads in manufacturing, healthcare, retail, banking and financial services. |
| E2E Networks | Provides GPU-powered cloud servers. NVIDIA cited Hopper systems connected using Quantum-2 InfiniBand. | Simulation, foundation-model training and real-time inference. |
| Netweb Technologies | Expands AI server systems for on-premises and hosted deployments. The 2024 announcement highlighted Tyrone AI systems based on NVIDIA MGX and GH200 Grace Hopper Superchips. | Dedicated infrastructure for organizations that need greater control over hardware, networking and data. |
These companies should not be treated as four brands selling an identical product. A managed cloud service, a hosted cluster and an on-premises server purchase have different pricing, availability, operating responsibilities and customer requirements.
Yotta and Shakti Cloud
Yotta’s Shakti Cloud is positioned as a sovereign Indian GPU cloud. NVIDIA later described it as offering access to H100 and newer Blackwell systems, including pay-per-use access. The exact current GPU inventory, quotas, pricing and availability require confirmation from Yotta; the public NVIDIA material does not establish that every configuration is continuously available to every customer.
Tata Communications
Tata Communications planned to combine NVIDIA accelerated computing with AI Studio and its network infrastructure. Its positioning is particularly relevant to enterprises that need managed connectivity and integration rather than simply renting an isolated GPU instance. NVIDIA’s 2024 announcement also said Tata planned to add Blackwell GPUs in the following year.
E2E Networks
E2E’s 2024 plan focused on GPU cloud servers and Hopper clusters using NVIDIA Quantum-2 InfiniBand. That networking is important for distributed workloads in which many GPUs must exchange data rapidly. NVIDIA’s later material describes an E2E Blackwell cluster on its TIR platform at L&T’s Vyoma Data Center in Chennai.
Rank #2
- Part number 900-53651-2500-000 and model: P3651
- This is the 2 slot version for when there is no empty slots between 2 slot cards. If you have one or more empty slots between the cards or the cards are 3 slot this NVLink will not work. See the attached images showing the card layout.
- NVLink 3.0 for any brand of RTX Ampere model graphics cards: 3090, A30, A40, A100 / H100 (Requires three NVLinks), A800, A4500, A5000, A5500, A6000
- This is the same as PNY part number: NVLAMP-2SLOT-BSP and RTXA6000NVLINK-KIT
- This is the same as Dell part number: 0RWJ7Y
Netweb Technologies
Netweb’s role is more closely associated with AI server systems for on-premises and hosted deployments. In 2024, NVIDIA highlighted its Tyrone systems using NVIDIA MGX and GH200 Grace Hopper Superchips. In later material, NVIDIA described Netweb-manufactured GB200 NVL4 systems using four Blackwell GPUs and two Grace CPUs.
Where Reliance and Jio fit
Reliance Industries was part of a separate NVIDIA partnership announced alongside the infrastructure news. The stated aim was to develop AI applications and services for Jio customers and build AI-ready data-center capacity in India.
NVIDIA said Reliance’s infrastructure could eventually expand to 2,000 megawatts. That is a long-term infrastructure ambition, not evidence that 2,000 MW was already operating in October 2024. The announcement did not provide a complete GPU count, deployment schedule or commercial structure. NVIDIA’s partnership announcement provides the stated scope.
What customers could use the capacity for
The infrastructure is intended to support more than general-purpose chatbots. Potential workloads include:
- training and fine-tuning large language models;
- Indian-language and multilingual AI;
- real-time inference and conversational agents;
- enterprise copilots and retrieval-augmented applications;
- healthcare imaging and drug-discovery research;
- financial-services automation;
- scientific simulations and visualization;
- industrial digital twins, manufacturing and robotics; and
- government, university and public-sector applications.
NVIDIA cited organizations including Sarvam AI, AI4Bharat, Qure.ai, InVideo AI, Assisto, Innoplexus and Zoho. These examples demonstrate reported or intended ecosystem use; they do not mean every provider offers the same models, GPU types, prices or access terms.
Why domestic AI compute matters to India
Indian organizations may want domestic infrastructure for several practical reasons:
- Data residency: sensitive business, health or public-sector data can remain within India’s jurisdiction, subject to the provider’s contracts and controls.
- Latency: serving models from India can improve response times for Indian users.
- Language coverage: local infrastructure can support research into India’s many languages and dialects.
- Availability: domestic capacity can reduce reliance on scarce overseas GPU resources.
- Startup access: cloud services let smaller companies rent compute instead of buying an entire cluster.
- Industrial policy: India can develop domestic models, applications, system integration and infrastructure instead of only consuming foreign AI services.
NVIDIA’s February 2026 material connects this ecosystem to the IndiaAI Mission, which NVIDIA describes as receiving more than $1 billion to support compute, sovereign datasets, frontier models, applications, education and trustworthy AI. That characterization should be understood as NVIDIA’s description of the program.
Rank #3
- Video/Sound Cards
- Passive Cooling
What “tens of thousands” and “180 exaflops” do—and do not—prove
“Tens of thousands” is a scale description, not one definitive public order total. It may refer to planned additions across several providers rather than one cluster. It also should not be confused with the number of GPUs installed, powered on, publicly rentable or available to a particular customer.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallThe nearly 180-exaflop figure is an aggregate performance claim. The cited announcement does not specify its numerical-precision basis. Real-world performance depends on GPU utilization, memory bandwidth, interconnect topology, storage speed, software optimization, model architecture and scheduling.
Accordingly, 180 exaflops does not mean that one customer can run a job at 180 exaflops, or that every application will receive proportional performance. Peak aggregate capacity is not the same as delivered training throughput or inference capacity.
The 2026 update: Hopper gave way to a broader Blackwell buildout
Later developments must be kept separate from the October 2024 announcement. In material published on February 17, 2026, NVIDIA said India’s AI cloud ecosystem included tens of thousands of NVIDIA GPUs and identified Yotta, L&T and E2E Networks among its cloud-infrastructure partners.
NVIDIA’s later account included:
- Yotta’s Shakti Cloud, described as powered by more than 20,000 NVIDIA Blackwell Ultra GPUs;
- an E2E Blackwell cluster on its TIR platform at L&T’s Vyoma Data Center in Chennai; and
- Netweb-manufactured GB200 NVL4 systems using four Blackwell GPUs and two Grace CPUs.
These figures and systems show how the story evolved toward newer GPU generations and a wider sovereign-AI strategy. They should not be retroactively presented as part of the original 2024 Hopper announcement. NVIDIA’s 2026 IndiaAI and infrastructure update is the source for those later claims.
Free tools Windows power users keep installed
One-click scans. No signup required.
Cloud access versus buying dedicated infrastructure
Cloud access is usually the better starting point when:
- workloads are experimental or variable;
- a startup needs capacity quickly;
- the organization lacks data-center and cluster-operations staff;
- capital expenditure must be minimized; or
- managed NVIDIA software and preconfigured environments are valuable.
Dedicated or on-premises systems make more sense when:
- GPU utilization will be high and predictable;
- data must remain under tightly controlled infrastructure;
- custom networking, storage or orchestration is required;
- long-term economics justify the capital expense; and
- the buyer can manage power, cooling, failures, drivers and scheduling.
A hybrid approach can keep sensitive training data local while using cloud capacity for bursts, disaster recovery or geographically distributed inference.
Questions buyers should ask
Domestic location does not guarantee immediate access or low prices. Before committing to a provider, an enterprise, startup or research team should request:
Rank #4
- CUDA Cores: 4608 / NVIDIA Tensor Cores: 576 / NVIDIA RT Cores: 72
- GPU Memory: 24 GB GDDR6 with ECC / Bandwidth: 624 GB/Sec
- System Interface: PCI Express 3.0 x16
- Four DisplayPort 1.4 Connectors
- 3D Stereo Support with Stereo Connector
- The exact GPU model, memory capacity and system topology.
- Whether the capacity is operational, reserved, hosted or still planned.
- Hourly, monthly and reserved pricing, including minimum commitments.
- Storage, data-transfer and egress charges.
- Availability, queueing and quota policies.
- GPU-to-GPU networking and storage performance.
- Region, data-residency, deletion and portability terms.
- Support service levels and hardware-replacement procedures.
- Whether NVIDIA AI Enterprise, NIM or other software licenses are included.
- Restrictions on drivers, kernels, frameworks and privileged operations.
The researched official sources do not provide dependable current prices or standard plans for the Indian offerings. Yotta is described as offering pay-per-use access, but the price, quotas and service terms must be obtained from the provider.
What remains unproven
The announcement does not answer several questions that matter to customers and investors:
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →- How many GPUs from the original plan are installed and operational today?
- What proportion is publicly rentable rather than reserved for large enterprises, government or strategic customers?
- Which systems use Hopper, H100, H200, Blackwell, B200 or GB200 hardware?
- What are the actual sustained training and inference throughput figures?
- What are the power, cooling and energy-efficiency characteristics?
- What prices, minimum commitments and wait times apply?
Those gaps do not invalidate the infrastructure expansion, but they limit what can responsibly be inferred from the headline numbers.
Sovereignty is not the same as technological independence
AI infrastructure located in India can improve data residency, latency and domestic access to compute. It does not eliminate dependence on NVIDIA hardware, CUDA-related software, global semiconductor supply chains, imported components, electricity, data-center operators or licensing conditions.
Nor do more GPUs automatically produce better Indian AI. Success also requires high-quality datasets, research talent, model-engineering expertise, evaluation benchmarks, privacy controls, safety processes and reliable production operations.
The Bottom Line
Bottom line: NVIDIA’s October 2024 announcement marked a major planned expansion of India’s NVIDIA-centered AI infrastructure: tens of thousands of Hopper GPUs across four initial providers, plus a separate Reliance partnership. It was an ecosystem buildout, not one publicly available supercomputer or proof of cheap, unrestricted compute. The 2026 Blackwell and IndiaAI developments indicate that the effort expanded, but customers still need to verify actual availability, pricing, performance and data-control terms provider by provider.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
