We need a moonshot for computing

A moonshot is required for computers.
The US government wants to advance the field of microelectronics research. Long-term competition, however, will necessitate embracing uncertainty

A study issued by the Obama administration in its last weeks sent shockwaves across the federal science and technology sector. Ensuring Long-Term US Leadership in Semiconductors was the title of the report, which issued a warning that the US was in danger of losing its competitive advantage in the chip business as traditional methods of chip construction collided with physical principles. Congress and the White House worked together to confront that possibility five and a half years later in 2022, when they passed the CHIPS and Science Act, a daring endeavor modeled after the Human Genome Project, the Apollo program, and the Manhattan Project. The US government has started preparing for the next era of computers throughout the span of three administrations

Gina Raimondo, the secretary of commerce, has even gone so far as to draw a direct comparison between President John F. Kennedy's 1961 demand to place a man on the moon and the passing of CHIPS. By doing this, she was recalling a US custom of arranging the country's innovation ecosystem to fulfill a bold technological goal that the private sector was unable to do on its own. Organizational issues and disagreements on the optimal course of action to guarantee national competitiveness in space existed prior to JFK's statement. Aspirations in technology tend to follow this pattern of developing on their own schedules

$11 billion will go into microelectronics research and development, and $39 billion will be used to bring chip manufacturing, or "fabs," and their major suppliers back to the United States. This was the amount allocated under the CHIPS and Science Act. The National Semiconductor Technology Center, or NSTC, would be the focal point of the research and development program. NSTC is envisioned as a national "center of excellence" that will bring together the brightest minds in the creative ecosystem to create the next generation of microelectronics

A year and a half later, chip fabrication facilities have begun construction in Arizona, Texas, and Ohio, and CHIPS programs and offices have been established. However, the CHIPS R&D initiative has the power to influence how the field develops in the future. In the end, national R&D objectives come down to two choices: either the US pursues real computer moonshots or it takes a cautious approach designed to maintain its edge for the next five years. Whether the United States plays it safe or goes "all in" on technology will depend on the NSTC's organizational structure and the programs it decides to undertake

Greetings on this day of reckoning
Gordon Moore, the late founder of Intel, famously foresaw in 1965 that the future of computers would require packing more transistors—tiny switches—onto flat silicon wafers. Moore predicted that the number of transistors would frequently double while the cost per transistor decreased, extrapolating from the invention of the integrated circuit seven years earlier. Moore, though, wasn't just speculating. Additionally, he was recommending a technological approach known as "transistor scaling," which involves shrinking transistors and packing them closer together to produce quicker and less expensive devices. This strategy not only ushered in the digital age but also gave rise to a $600 billion semiconductor industry

Moore, who was always perceptive, did not anticipate that transistor scaling would be permanent. He called this "day of reckoning," the moment at which this process of shrinking will hit its physical limits. If it hasn't already, the chip industry is currently very near to arriving at that day. Technical difficulties are growing and costs are soaring. According to industry road maps, transistor scaling may not hit its physical limits for another 10 to 15 years, and it may cease to be lucrative before then

In the short term, the semiconductor industry has implemented a two-pronged approach to maintain chip advancement. In order to accelerate processing, on the one hand, "accelerator" chips are being built for particular uses (such AI inference and training). On the other hand, in an effort to cut costs and increase customizability, businesses are creating hardware using smaller functional components known as "chiplets." You can assemble these chiplets in a stack or side by side configuration. The 3D method may be a particularly effective way to increase speeds

Over the next ten years or so, this two-part plan will be helpful, but its long-term effects are limited. It still uses the same transistor-building process, which is rapidly nearing its end, for starters. We will also still have to deal with energy-hungry communication bottlenecks even with 3D integration. It's unclear how long this strategy will allow chipmakers to create computers that are both more competent and less expensive

constructing a moonshots institutional home
Creating alternatives to traditional computing is undoubtedly the better course of action. Numerous contenders exist, such as quantum computing, neuromorphic computing—which replicates brain functions in hardware—and reversible computing, which could potentially exceed the physical bounds of computer's energy efficiency. Furthermore, a plethora of innovative materials and technologies, including superconductor electronics, magnetic materials, and silicon photonics, may be employed to construct computers in the future. Hybrid computing systems might potentially be created by combining these capabilities

None of these promising technologies are new; scientists have been working on them for a long time, and the private sector is undoubtedly seeing advancements in quantum computing. However, only Washington can provide the R&D funding and convening authority to enable these innovative systems to grow to scale. Microelectronics innovations have typically come about piecemeal. However, to fully realize new computational approaches, a completely new computing "stack" must be constructed, starting with the hardware and working its way up to the algorithms and software. This calls for a strategy that can unite the whole innovation ecosystem around distinct goals in order to address several technical issues concurrently and offer the kind of assistance required to "de-risk" otherwise hazardous endeavors

Is it wiser to make large wagers on future breakthroughs or to concentrate on increasing competitiveness in the short term?

These initiatives can be led by the NSTC. It would be wise to concentrate on moonshot projects in order to follow DARPA's example and be successful. It will be necessary to shield its research program from external influences. It must also support visionaries with a sizable internal technical workforce, including program managers from academia and business

The center's investment fund must also be carefully managed, taking inspiration from top deep-tech investment funds now in existence. Examples of these practices include providing entrepreneurs with access to resources like tools, facilities, and training, and guaranteeing transparency through due diligence procedures

The NSTC is still in its infancy, and success may come after a long and convoluted route. However, this is a critical juncture for US leadership in microelectronics and computing. We'll need to consider carefully what kinds of institutions we'll need to bring us there as we map out the future for the NSTC and other R&D goals. We might not have another opportunity to do it correctly