Going back to 2006 when you founded the company, what did you and Martin see that made you bet the company on microinverters when the entire industry was built around string inverters?
I came primarily out of communications, doing cutting-edge fiber-optic technology during the internet boom. What we realized was that decentralized or distributed architectures always win in the long run, and they win for cost, performance and reliability. The proof points are everywhere, mainframes to PCs, ringing telephones to smartphones, data centers built not as big monolithic compute but as thousands of replicable blade servers. In network parlance, intelligence gets pushed deeper into the network, endpoints become very intelligent, they make autonomous decisions and optimize locally while staying hyper-networked and globally aware. When we started looking at the energy infrastructure, over 125 years it really hadn't evolved. It stayed very central, massive generation plants, transmission, substation distribution, and the home is just a receiver of energy.
We said if technology history is correct, that needs to evolve into something completely decentralized, and applying first principles, the home becomes the unit of intelligence. It produces its own energy, stores its own energy, uses it intelligently, and participates in the optimization of the wider infrastructure.
That was the genesis. You don't try to boil the ocean, so we started with power management. We asked why power conversion for solar is being done centrally. It needs to be completely distributed, and that's when we started thinking about microinverters. But here's the key, microinverters aren't about shrinking a central or string inverter into something small. It was about using semiconductors and software to re-architect the device from the bottom up, relying on our own custom semis, advanced wide-bandgap technologies like silicon carbide and gallium nitride, and making it a software-defined system. We built a power management platform whose applicability is very broad. It started with solar, we added the same technology to our batteries, then to EV chargers, including the bidirectional EV charger. The technology never stops. We're now developing a solid-state transformer for data centers using that same platform.
With the 25D credit having expired at the end of 2025, how deep is the demand hole in 2026, and when do you expect US residential to find its floor?
Our view is that we are very close to the floor or at the floor. 25D was a specific tax credit for residential cash and loan customers, and 80 to 90 percent of systems in the US are financed. However, there is another financing tool now called prepaid lease, and prepaid lease with loan, that is a good proxy or substitute for 25D. In addition to traditional lease, the homeowner can prepay all the lease upfront and own the system after about five years. So it's a good option to continue financing the system and capture the 25D equivalent tax benefits. We're seeing very strong adoption of that financing package, so I think things are beginning to recover.
Here's my personal view. Incentives need to be a catalyst, not a crutch. Unlike the fossil fuel industry, which has been incentivized for 200 years, I'm a big believer in free markets. Government should not be in the business of picking technologies, incenting one and de-incenting another. Let the best technology win, the most cost-effective, safe, clean and easily deployable technology. But I'm pragmatic. We can't be dogmatic and say only this technology should be deployed. The plane is flying and it needs to be serviced now. Data center demand growth is unprecedented, onshoring and industrialization growth is unprecedented, and we cannot and should not stop that, because that hurts the economy and works against affordability. We've got great technology today that interconnects very quickly, solar, batteries, EV chargers. Let's go for it.
Given how you see the technology evolving, when the 45X manufacturing credit starts to step down later this decade, does your US manufacturing footprint still make sense without that subsidy?
100 percent. Technology is always a point-in-time discussion because it's constantly evolving, and we release products based on a platform roughly every 18 months. That's our cadence. Every 18 months we release a new solar inverter, a new battery, a new EV charger, a bidirectional EV charger. IQ8 was our previous-generation product and we're now migrating to IQ9, which is based on a brand-new ASIC called Kestrel as well as gallium nitride. That puts us on a whole new trajectory in terms of cost and performance. It's not just about the product itself, it's about manufacturability. We continually push more automation so we can build more product, because demand is unprecedented.
I come back to the original notion that incentives need to be catalysts, not crutches. I've learned that lesson over 20 years. Innovate or die is a standard Silicon Valley mantra, and innovation isn't just releasing new features, it's implementation, business-model innovation, product, cost, manufacturability, everything. Only the paranoid survive, as Andy Grove put it. That's built into our DNA. So when 45X steps down, we'll already be at a structurally lower cost point and the manufacturing footprint will still make sense on the merits.
Microinverters were a radical architectural bet in 2006 that the industry hadn't caught up to yet. What's the equivalent bet you're making today?
That same conviction in distributed architecture runs through virtually everything we do. Take batteries. The cell is only a very small part of the battery. What matters is battery management, the charge converter, inverter, software, packaging, thermal management. We bet on all of that. We don't make solar panels, that's commodity, and we don't make cells, photovoltaic or chemical, that's completely commodity. We do everything in between, which is where all the intelligence sits. Now in the world of AI and agentic systems, we've embedded an AI agent in the homeowner's app, so you can ask any question and it looks at the state of your system and tells you exactly what's going on. We forecast your production for the next day, forecast your consumption, and do real-time optimization, when should you charge the battery, when should you discharge, when should you buy from the grid, when should you sell. Everything is dynamic, and completely transparent to the homeowner.
The other big bet is what we're doing for AI data centers with the solid-state transformer. SSDs apply not only to data centers but to large-scale battery storage, fast DC charging, the lot. Most of the SSD work in academia and now in industry still takes a very centralized approach. We took a supercluster approach. Instead of building an SSD with five or six or eight power modules, we are building one with 342 power modules. It's a supercluster, exactly like the way NVIDIA stacks GPUs with NVLink to create a supercomputer. We have a supercluster of power modules to build a 1.25 megawatt SSD. You have to be convicted on what you believe in, and for us decentralization and distributed architectures always win for cost, performance and reliability. The beauty of the platform is that you don't need every element. You can have a home with a bidirectional EV charger only, solar only, battery only, or like me, with everything. Across 100 million homes, the right combination of solar, battery and EV charging matters, and the technology is here today. The US sits at 4 percent residential deployment. It's embarrassing. Germany is at 15 to 18 percent, the UK significantly lower, France in the teens. 80 to 90 percent of the work is still left to do. The technology to address affordability, data centers and grid hardening is already here, and Enphase is doubling down on it.
