Hyperscaler interest in fusion has reached a fever pitch in recent months, as tech giants strike power-purchase agreements and strategic investments across multiple fusion startups. Even so, the push is generally considered a play for speculative, far-off returns.
But efforts to shore up a fusion supply chain could have a much nearer-term impact on a different energy problem: speed-to-power for data centers.
Fusion relies on high temperature superconducting tape, known as HTS, which can carry very high currents with almost no electrical resistance once cooled. This allows more power to flow through smaller, lighter cables.
For years, the technology was only viable in specialized use cases like fusion, because it was hard to find and required long lead times. Commonwealth Fusion Systems, for example, created an HTS supply chain from scratch to meet its needs, because the tape had never been manufactured at scale before.
But the boom in fusion has changed things. According to Husam Alissa, director of systems technology at Microsoft, scaled-up production has driven process improvements, cost reductions, and better quality control. And that has made it viable for data center use — and specifically for Microsoft, which has emerged as the tech’s hyperscale champion.
The tech giant is hopeful that HTS could address expansion challenges both inside and outside of data centers, by reducing the amount of space required to move power. Microsoft has spent the last several years doing extensive research and prototyping, and is now working to kickstart a wider embrace of the technology, encouraging utilities and other hyperscalers to adopt and de-risk it.
“Inside the data center, the most interesting use case for us is high density power delivery closer to the rack,” Alissa explained. In traditional architectures, electricity is typically stepped down from high-voltage transmission to low-voltage delivery at the very end of the line because low-voltage power loses significant energy to resistance in copper cabling. HTS, however, can carry massive currents at low voltage over longer distances with virtually no loss.
Alissa pointed to Microsoft’s work with superconductor startup VEIR, which the tech giant has supported through its Climate Innovation Fund. VEIR originally targeted long-distance transmission lines for its technology, but has pivoted to the more immediate AI data center use case. In November, Microsoft and VEIR conducted a high density power delivery pilot, in a simulated data hall, delivering 3 megawatts of power via a single HTS cable.
“We were able to do that with more than 10x reduction to cable dimensions and weight of conventional methods,” Alissa explained. “As AI rack power goes up, traditional cabling could get bulky, could get complex. Deploying HTS opens up more flexibility in designing data centers, he added, pointing to increased flexibility in rack layout, electrical room locations, and relaxed distance constraints.
On the grid side, Microsoft is hopeful that HTS can offer a hack when it comes to community opposition and land-use hurdles. Traditional transmission corridors can require rights-of-way up to 70 meters wide to accommodate massive towers. Because HTS cables are up to ten times more power dense than traditional cables, and can be buried in much narrower trenches, they could open up geographies that are currently challenging for data center development.
The “anchor” of HTS benefits, Alissa said, is the potential to avoid community opposition, which continues to impede data center development even as creative power deals seek to get chips online faster.
“It means less land use, less visual impact, less disruption, and still having access to power,” Alissa said.
For more on Commonwealth Fusion Systems’ creation of an HTS supply chain, listen to CFO Rick Needham’s interview on The Green Blueprint:
The path forward
The hurdles now facing HTS are less to do with science, and more to do with systems and adoption, Alissa said. And on that front, it’s still early days.
Microsoft and its partners still need to figure out exactly where in the data‑center architecture HTS provides the most leverage, he said, and continue encouraging a resilient supply chain and business model for the tape itself.
Inside the data center, there’s the challenge of operating a high-availability cryogenic plant, ensuring the cryogenic cooling system can meet the “five nines” reliability standards of a modern cloud facility without adding operational complexity or risk for technicians.
On the grid side, utilities and operators have to agree on standards and interfaces — and overcome the “psychological barrier” to deploying unfamiliar transmission technology at scale.
There isn’t a clear deployment timeline for either use case, nor is it obvious which path would pick up the most traction among hyperscalers, Alissa said. Importantly, Microsoft hasn’t itself decided whether HTS will provide the most return on investment for use in greenfield sites with access challenges, or expanding capacity at existing data centers.
“I don’t think we have complete visibility to that yet…there could be opportunities in both,” Alissa said. “It’s really [about] nailing down those bottlenecks and then matching the technology with a problem.”
