The energy industry is in the first stages of a once-in-a-century transformation. And one of the most important aspects of this shift is that EVs, solar farms, grid equipment, and appliances will inherently rely more on digital technologies. As Hamed Heyhat, General Manager of Grid Automation at General Electric, says, “decarbonization cannot happen without digitization of the grid.”
So how will we see that impact roll out in the chip industry?
1. The grid becomes a network
Power grids have been called the most complex machines ever built. Utilities deliver power to millions of customers spread over thousands of square miles in real-time on a 24/7 basis.
But flexible they aren’t. Utilities have little visibility into how much power you consume and even less control over how much you use. For a margin of error, they invest in expensive, little-used peaker plants. California alone has 17GW of peaker plans and they are used less than 15 percent of the time.
Transforming the grid into a smart, multi-directional network – where finessing power loads with AI and processing replaces the brute force of excess capacity– is already underway. Span has created a smart electrical panel powered by a quad-core Arm processor and AI that effectively turns a home into a microgrid: residents can run their home from a car during natural disasters or shut off appliances remotely to save money.
Enphase, meanwhile, has developed its own Arm-based ASICs for powering microinverters to increase solar output, controlling battery systems and selling excess power to utilities. In New England, utilities are offering customers up to $1,000 a year for letting them have limited access to battery systems on select days. Inverters and optimizers “smartened up” with CPUs and software have gone 30 percent to 80+ percent of the U.S. residential market.
Demand for visibility grows exponentially as you move to distribution and transmission networks. Edge Impulse is experimenting with computer vision at the edge to spot faults and fires more quickly while Awesense, is building renewable microgrids for corporate campuses. It has set a goal of avoiding 100 million tons of CO2 by 2025. Because of the stringent safety requirements in this market, one can anticipate that some of the design ideas for isolating workloads and/or ensuring reliability integrated into ADAS systems for cards will increasingly find their way into these devices.
2. So do buildings
Critics might argue that outfitting every home with a quad-core CPU won’t fundamentally boost shipment volumes. There are only around 140 million residences and nearly 6 million commercial buildings in the U.S., putting the TAM for smart building controllers in this country at 150 million.
But smart building systems have to talk to something. Appliance manufacturer Arcelik has found that running relatively simple AI algorithms at the edge can reduce refrigerator power consumption by 10 percent. Deployed across Europe, this could replace nine power plants. Aquaseca is developing a device that uses sound recognition to detect pipe leaks: insurers are testing devices like this to put a dent in the annual $13 billion worth of water damage claims. The space, power and thermal constraints of household appliances also make intriguing testbeds for NPUs.
There will be also a flotilla of less intelligent, less programmable, but nonetheless “aware” devices like light bulbs (40 per home) and electrical outlets (75). The volumes of MCUs, flash, embedded communications needed in buildings will reach billions of units while the sophistication of the embedded intelligence will grow over time.
3. Silicon boosts renewables
Most of the costs of solar and wind facilities occur upfront during the construction of the plant. The largest variable cost factors are downtime and maintenance.
At wind farms, the main culprits are the bearings moving the nacelle and blades. Running a turbine to fail can cost $150,000 worth of repairs and lost production with the turbine often going back to the shop. Pre-emptively repairing them in situ via IoT and 5G or LTE runs $5,000.
While there are few if any moving parts in solar fields, pre-emptive repairs via IoT and AI could boost profits at a 50MW solar plant by $500,000 a year, or $17 million over its lifetime. (Added bonus: the CPU in the inverter can be leveraged for many of these tasks). Utilities such as Arizona Public Service are managing solar plants spanning thousands of square miles with small crews because of this sensory awareness. There is no shortage of data for AI algorithms to analyze: a 1MW wind farm will produce 9x more data than its fossil counterpart while a solar plant will generate 40x more.
How big is the opportunity? The IEA estimates that we will need $4 trillion a year in clean energy investment starting in 2030. Solar and wind are expected to grow from 10 percent of worldwide capacity now to 70 percent by 2050.
4. The not CMOS industry charges up
Battery packs have plunged in price from around $1,200 per kilowatt/hour to $135 over the last decade. A good proportion of the gains came through the magic of volume manufacturing. Reducing the price to $45 per kwh by 2035 will require re-engineering storage.
Silicon Carbine (SiC) and Gallium Nitride (GaN) semiconductors can increase power conversion and efficiency of inverters and SoC. In turn, that translates to more miles for an EV, more hours of off-grid power in a home, and/or smaller, more affordable battery packs. GaN and SiC can also accelerate EV charging or reduce transmission and distribution losses on the grid. Deployment gets simplified as well: an SiC-based solid-state transformer that would fit into a suitcase can replace 8,000 pounds of copper wire inside a traditional transformer.
Sic and GaN have existed for years, but the volume demand has been comparatively low. A switch in fuel source changes the picture. Ford and other manufacturers reportedly are signing deals to lock up capacity.
5. Design and manufacturing get clean
The push toward sustainability will also mean that the semiconductor industry itself will have to become more energy efficient. Last year, TSMC signed an agreement to obtain 920 MW of offshore wind for running fab operations. The company is now one of the ten largest purchasers of renewables in the world. (And will intermittency be a problem? Bloomberg New Energy Finance estimates that spare storage capacity in UPSes could provide gigawatts of capacity to the grid.)
Meanwhile, we will all find ways to reduce the time, energy, and emissions involved in design, test, and verification. Arm, for instance, has begun to shift many design processes to the cloud. We anticipate this will help us reduce data center energy by 45 percent while accelerating tasks by 6x in some instances.
The long and winding road
The power and utility industry is somewhat different than many traditional electronics markets. Technology investments often have to be approved by public commissions. Technology trials can take years. The heightened concern for safety also requires suppliers to meet stringent requirements. But we do know this: the change is coming and Arm, together with the rest of our industry, will play a large part in it.”
Report: Decarbonizing Compute
How the Arm Ecosystem Is Laying the Foundation for a Net-Zero Emissions Future
This blog originally appeared on Semiconductor Engineering
Any re-use permitted for informational and non-commercial or personal use only.
Brian Fuller and Jack Melling