Imagine you’re the next multi-billion dollar AI company, ready to crack consciousness or mine the latest crypto coin. You just dropped eight figures on signing bonuses for talent, your GPUs are firing, and your models are queued. On paper, you’re ready to scale…that is, until an operator for a decades-old utility tells you it’ll be years before you can plug in.
Welcome to the AI boom’s dumbest bottleneck, where the world’s fastest-growing technology is chained to one of the slowest-moving systems: the grid. These companies aren’t exactly known for their patience. And right now, new data centers worldwide are facing multi-year wait times. Texas expects 205 gigawatts of load queued up by 2030, according to ERCOT’s Kristi Hobbs. To put that in perspective, California’s total installed electric generation capacity is below 100 gigawatts.
This isn’t just an American pickle. Global interconnection queues are now so clogged that some cloud providers are paying up to 10 times the cost to skip the line. Imagine bribing your way into a wall socket.
That’s where Emerald AI, the latest name on the Lowercarbon roster, steps in. The company is led by Dr. Varun Sivaram, a Rhodes Scholar with a PhD in physics and former Chief Strategy Officer at Ørsted and CTO at ReNew Power.
Varun and his team are building the software layer that turns data centers into flexible grid allies. This transforms AI’s parasitic loads into symbiotic ones that support the growth of our energy systems. In doing so, Emerald is solving a problem that suddenly seems to be on the tip of everyone’s tongue: speed to power.
Today, we are excited to announce our lead in Emerald’s $18m Seed Extension, alongside Nvidia, Radical Ventures, Salesforce, National Grid Partners, Amplo, Earthshot, Trust, FYSK, John Doerr, and other partners who are helping Varun turn that capital into electrons on the wire.
A Gem of an Approach
Here’s the thing about the current grid: it was designed for predictable loads, like factories running eight hours a day and laundry machines spinning at dusk. Now we’re asking it to handle never-before-seen levels of compute every second of the year with zero flexibility and zero warning. Add an entire state cranking their ACs on full blast to compensate for how damn hot it is outside, and it’s clear this rusting infrastructure is simply not up to snuff.
Until the grid joins the 21st century, which is on track to happen in the 22nd century, Emerald is helping utility companies allocate electrons with far more efficiency. When the grid is tight, they can pause, shift, or stagger compute jobs across clusters for minutes at a time, without breaking performance guarantees. When supply loosens, workloads automatically ramp back up. It’s invisible to the data center and a business saver for AI companies needing to skip the line to get online before their next investor meeting.
With that flexibility, utilities can do something they’ve never done before: green-light interconnection without waiting for costly new substations or transmission lines. Because Emerald-equipped facilities dial demand up or down in sync with grid conditions, they count as “controllable loads” rather than fixed liabilities. That changes the math. Instead of overbuilding for the worst 10 minutes of the year, utilities can connect more megawatts today. The result? Years shaved off project timelines—and tens of billions saved on concrete, copper, and consultants.
By letting compute follow the rhythm of supply, Emerald converts stranded capacity into usable watts. A Duke University study found that just a few hours of flexible throttling could unlock the equivalent of every U.S. data center currently in the pipeline. Turns out, the grid’s real constraint isn’t energy, but timing.
The World’s First AI Factory
Emerald is building the Aurora AI Factory in Virginia together with NVIDIA, Digital Realty, Dominion Energy, PJM, and EPRI. This will be a 96-megawatt, power-flexible facility designed to demonstrate how AI can grow in sync with the grid. Think of it as a model home for the next generation of bit barns.
When this model takes hold, it will unlock up to 100 gigawatts of hidden capacity in existing networks, equivalent to powering every data center planned for the next decade, without requiring a single new power plant.
The Electric Power Research Institute also selected Emerald as one of ten national demonstration projects to define the first flexible interconnection standards in the U.S. That’s utility speak for: “We trust these guys not to blow anything up.”
Across the Atlantic, Emerald and National Grid are launching the UK’s first live trial of a grid-aware AI data center that dynamically adjusts consumption to local conditions—speed to power, powered by flexibility. Same principle, different accent.
Code Scales Faster than Interconnection
AI infrastructure today is rigid, overbuilt, and blind to grid conditions. That approach isn’t scaling. The next generation of AI factories will need to think more like the models they train: adaptive, self-tuning, and responsive, with a little less hallucination all around.
By embedding flexibility into the core of compute operations, Emerald offers AI infrastructure teams a faster, cheaper, and cleaner path to scale.
The reference design they’re co-developing with NVIDIA and EPRI could become the template for how utilities and hyperscalers plan new capacity. If it works, flexibility won’t be a retrofit. It’ll be the foundation. And once that happens, concrete will start taking notes from code.
The Watt Rules AI
We invested in Emerald because the next frontier in AI won’t be about just more chips or bigger models. It may come down to who gets power first. Varun’s mix of policy, utility, and tech experience makes Emerald uniquely credible in a market where execution depends on trust.
In venture-speak, Emerald could become the control plane for flexible compute—the Stripe or Snowflake of the power grid. In human-speak, they ensure AI companies get their fill while leaving more than enough watts for the rest of us.
As always, if you’re working on something this ambitious and high-impact, come see us. We’ll leave the light on for ya.
The post The Express Lane for Electrons appeared first on Lowercarbon Capital.
