SIP
Efficiency Is Existential: On Building a New Magnitude of Infrastructure
Even before today’s AI era, energy and compute demand were rising at a pace that should have made everyone alert. The electrification of seemingly everything has been underway for years. Data consumption has been surging. Digital services have been embedding themselves into every sector of the economy. The strain on our underlying infrastructure was building long before ChatGPT ever made headlines.
That is the backdrop against which we founded SIP. Shortly after spinning out from Alphabet in 2019, we began building businesses at the convergence of energy and digital infrastructure because we recognized that these systems were becoming the gating factor for economic growth. AI has not created that reality. It has exacerbated it and made it impossible to ignore.
What has changed most dramatically in the past few years is the sheer magnitude of infrastructure to be built.
In 2026 alone, the four largest U.S. hyperscale tech companies are projected to spend a combined $650 billion in capex. Just six years ago, in 2020, that same group spent approximately $94 billion, a fraction of this year’s figure. In parallel, U.S. utilities are planning to invest more than $1.1 trillion in grid upgrades between 2025 and 2029, a level of spending roughly equivalent to what they deployed over nearly twice as many years from 2015–2024.
To put it plainly, we have entered a radically different league of investing in and developing energy and digital infrastructure. And in this era, our industry’s historical infrastructure development playbooks will not be remotely sufficient. The unprecedented magnitude and velocity present new questions and demand new answers of infrastructure developers.
There is the question of scale: How will we deliver enough energy and compute capacity, and urgently enough, to match exponential demand?
There is the question of affordability: With forecasted investments reaching unprecedented levels, particularly with hyperscalers’ planned infrastructure spend comprising a substantial portion of their balance sheets, how do we make this buildout cost-effective?
And there is the question of resilience: As compute and energy requirements rapidly evolve, how do we ensure that the infrastructure we design today is not obsolete by the time construction is complete?
Meeting these requirements—scale, affordability, and resilience—requires new approaches. This buildout of energy and compute infrastructure, arguably the largest infrastructure expansion in our country’s history, must be undertaken with efficiency at its core.
Efficiency Is Existential
In prior infrastructure cycles, when energy and compute infrastructure was built one megawatt at a time, inefficiency was tolerable. If utilization fell short or the flexibility of a system was limited, the impact was manageable. Developers, utilities, and customers could absorb the costs. The system would survive.
But as we onboard hundreds of gigawatts, that calculus breaks down. When annual capital deployment is measured in hundreds of billions of dollars, percentage points of inefficiency are not rounding errors. They determine whether the system holds or fractures.
That is what we mean when we say efficiency is existential. For household ratepayers, inefficiency means electricity becomes entirely unaffordable as utilities socialize the massive cost of underperforming assets across the customer base. For hyperscalers, it means eroded returns, increasing the risk of stranded infrastructure and putting pressure on balance sheets already stretched by unprecedented capex commitments. For utilities, it means investing billions into assets that fail to fully deliver, undermining reliability and inhibiting their ability to serve growing demand. At gigawatt scale, inefficient infrastructure does not merely underperform. It pushes our systems and stakeholders to the brink.
The path forward is clear: a more innovative focus on efficiency. We must both utilize the infrastructure that already exists more intelligently, and when we build new infrastructure, we must design it to adapt to the changing operating environments it will service over its 30– to 50–year life. The only way to deliver the scalable, affordable, and reliable infrastructure that this moment and the future requires is to build it far more flexibly than we have in the past.
Efficiency can no longer be a concept at the margin. It is table stakes for developers, utilities, and customers alike.
It is also core to how SIP builds infrastructure.
This Is What We Do
We founded SIP with this premise: society's potential is gated by the infrastructure systems that support it. In our early years, we pursued opportunities across multiple verticals. But as we deepened our understanding of where the need was greatest, energy and compute infrastructure took primary focus. Access to affordable, clean, and reliable energy and compute—two of the most essential inputs for equitable economic growth and advanced innovation—is integral, particularly in an increasingly competitive global landscape.
Today, SIP is deliberately concentrating its capital and operating strategy at the intersection of flexible energy systems and grid-interactive compute infrastructure.
We recognized years ago that legacy energy and compute systems were unprepared for rising demand, and we began delivering solutions. Our approach centers on building businesses that maximize the productivity of existing systems as well as businesses that develop new, intrinsically-efficient infrastructure that is flexible from day one.
Through Renew Home, our virtual power plant (VPP) subsidiary, we demonstrate efficiency maximization at scale through the orchestration of residential energy systems. In partnership with 110+ utilities across the country, our VPP coordinates the energy usage of nearly 7 million households nationwide to provide over 4 GW of flexible capacity to the grid—the equivalent of two Hoover Dams worth of dispatchable load. When summer heat waves push the system to its limits, our network responds in seconds, shifting demand and stabilizing supply before emergencies become blackouts. The U.S. Department of Energy estimates that scaling VPPs could meet up to 20% of peak demand by 2030, saving $10 billion annually in avoided infrastructure costs. This is a solution that works at scale today, and SIP is proud of building one of the largest VPPs in North America to meet the moment.
Through our subsidiary Verrus, we develop flexible data centers that are designed with efficiency at the core. In collaboration with the National Laboratory of the Rockies, we demonstrated the first full-scale simulation of a grid-interactive data center that can curtail 100% of load within one minute of a utility signal, without impairing customers' server uptime requirements. Our designs integrate long-duration battery storage that provides grid services during normal operations and backup power during outages. They use 99% less water than standard facilities. Flexibility is built into the architecture from the start. Data centers have been framed as a burden on the energy system. We are proving they can be an asset to it.
This is what building smarter looks like in practice: unlocking latent capacity where it already exists and embedding flexibility into new infrastructure from the outset.
Where The Market Is Getting It Wrong
In this current market, however, not all infrastructure developers are building what this moment and the future require.
As trillions of dollars pour into our energy and digital infrastructure space, we are witnessing an industry gold rush. Opportunistic entrants are staking claims. Bold projects with ambitious valuations are sweeping the news. Yet much of the market is still building as if this new magnitude of infrastructure has not eviscerated the old rules. Many builders, both new and incumbent, are taking inefficient and outdated approaches.
Across the data center sector, developers without an understanding of the need to build in flexibility are constructing rigid facilities that risk becoming obsolete in just a few years as compute requirements rapidly evolve. Many cannot even cogently explain what designs the market will require—such as size, load profile, locations, architecture—when we start running more and more integrated inference and AI training loads. Everyone wants exposure to AI-driven demand, but few know how to build what is actually needed to accommodate it.
On the grid side, the traditional, ongoing response to rising demand has been to expand transmission and distribution infrastructure. While that approach served the industry well during periods of steady, predictable growth, today’s demand trajectory is less linear and more volatile. In this environment, expanding physical infrastructure without first maximizing existing asset utilization leads to inefficient capital deployment. As an industry, we must urgently and systematically elevate proven, innovative technologies, such as advanced load coordination, distributed energy resources, storage integration, and improved forecasting, from supplementary programs to core planning principles.
Our reflection on the market is this: the participants who are approaching this as a traditional infrastructure cycle are misreading the moment. They will also not endure. Credit markets are already starting to reward assets that are flexible, efficient, and durable, and punish those that are not.
Far from a temporary spike, what we are experiencing is a structural shift in the magnitude and velocity of energy and compute demand and invested capital. Projects underwritten on assumptions of steady load growth, static utilization, and passive grid integration will struggle in an environment defined by volatility, rapid technological change, and tightening capital access. Those who are not prepared to build scalable, affordable, and resilient systems will not survive.
The Path Forward
How efficiently our society builds energy and digital infrastructure over the next handful of years will shape outcomes on economic progress, equitable growth, energy security, leadership in AI and innovation, and climate for generations to come.
This buildout will involve considerable capital, but it also requires much more than capital. It requires precise, novel coordination of supply and demand. It requires the integration and orchestration of innovative software, hardware, and real-world energy assets. It requires new generation, new transmission, and new storage, and it requires making far better use of what already exists.
SIP is structured to deliver all of the above, not in theory, but in reality. We build the businesses that enable the next generation of critical energy and digital infrastructure systems. We have been doing so since before the modern AI era, and we will continue delivering solutions that address today’s needs and tomorrow’s evolving demand.
Efficiency has long been foundational to developing scalable, affordable, and resilient infrastructure.
Now, at gigawatt and trillion-dollar scale, that efficiency is existential.
Brian Barlow
CEO, Sidewalk Infrastructure Partners
Related Insights
