OpenAI’s 30GW AI plan: The biggest compute race yet
OpenAI plans 30GW of AI compute by 2030. Here’s what it means for data centres, energy, chips and India’s AI ecosystem.
AI is now limited by electricity. OpenAI has set a target to scale its computing capacity to 30 gigawatts by 2030. The plan, reported on 23 April 2026, signals a massive shift in how AI systems will be built and powered in the coming years.
To put that in perspective, this is not just about more servers. It is about building infrastructure at the scale of entire power grids. Here's everything you need to know about it!
From gigabytes to gigawatts
Until recently, AI discussions focused on data, models and algorithms. Now, the conversation is moving towards energy. OpenAI’s roadmap shows rapid growth. The company moved from around 1.9 gigawatts of compute at the start of 2025 to a near-term goal of 10 gigawatts.
Of that, nearly 8 gigawatts is already close to being secured. The 30 gigawatt target takes this to another level. It represents a shift from scaling technology to scaling infrastructure.
Why power is now the real bottleneck
Training and running advanced AI models require enormous amounts of energy. Every query, every model update, and every deployment consumes power. As models grow larger and more complex, this demand increases exponentially.
Leaders like Sam Altman have acknowledged that the future of AI depends on a reliable energy supply. Without it, even the most advanced systems cannot operate at scale. This makes energy the new limiting factor in AI growth.
What does it take to build 30GW?
Reaching 30 gigawatts is not just a technical challenge. It is an infrastructure problem. OpenAI will need to invest in:
- Large-scale data centres designed specifically for AI workloads
- Stronger grid connections to handle continuous power demand
- Long-term partnerships with energy providers
- Potential access to dedicated power generation
In some cases, this could mean building or co-locating with power plants to ensure uninterrupted supply.
Cooling is another critical factor. High-density AI clusters generate significant heat, requiring advanced cooling systems to maintain performance and reliability.
The semiconductor challenge
Power is only one side of the equation. The other is hardware. AI systems rely on specialised chips, including GPUs and custom accelerators. OpenAI is reportedly working on its own chip designs to improve efficiency and performance.
However, supply constraints remain. High-bandwidth memory, a key component for AI workloads, is still limited despite increased production from companies like Samsung and SK Hynix. Scaling to 30 gigawatts will require consistent access to these components at massive volumes.
A race that others are also running
OpenAI is not alone in this push. Tech companies and cloud providers are all racing to expand AI capacity. However, a 30 gigawatt target places OpenAI among the most ambitious players in the space.
If achieved, it could redefine the baseline for AI compute. This would enable faster model training, more advanced capabilities and quicker product cycles.
Why India should pay close attention
For India, this shift brings both challenges and opportunities. As local companies build AI products, access to reliable power and infrastructure will become critical. Data centre expansion, grid resilience and energy policy will play a bigger role in shaping the ecosystem.
There is also a need for skilled talent. Building and operating large-scale AI infrastructure requires expertise across engineering, energy management and operations. This could open new opportunities in the Indian job market.
At the same time, financing models that combine renewable energy, storage and stable power sources will become increasingly important.
What to watch next
The immediate focus will be on execution. Can OpenAI secure its near-term 10 gigawatt target? Will it establish long-term energy partnerships? How quickly can it scale custom chip production? These are the milestones that will determine whether the 30-gigawatt vision becomes reality.