| | | | | The semiconductor industry got a jolt last week from an unknown startup named Substrate claiming it could take on two of the most entrenched players in global technology.
Substrate told us for our story published last week that it has built a new kind of chipmaking machine that’s as good as, if not better than, the latest extreme ultraviolet lithography tools manufactured by ASML Holding NV, the world’s top chipmaking equipment company. EUV rigs are so highly prized that each costs hundreds of millions of dollars and their sale to China is prohibited.
As if rivaling ASML wasn’t enough, the startup said it plans to run its own US-based chip factory, putting it in direct competition with Taiwan Semiconductor Manufacturing Co., which dominates contract chip production. The effort is billed as a patriotic play for the US to reclaim leadership in a technology that was born in Silicon Valley, but is now controlled by faster-moving foreign players.
Startups in this part of the world regularly claim they’re going to change the world. It’s almost a box they have to tick. They’re usually given an extended grace period to see how that plays out. But this one is already facing questions that are as rigorous and specific as its plans might be grandiose.
Compared to consumer technology or social media, media platforms and even software, the chip industry remains an unforgiving place where progress is hard-won, but very easy to measure.
Substrate claims it has perfected already-tried and abandoned techniques — in this case X-rays — to do the work of burning patterns into materials deposited on silicon wafers. That means its machine would be able to replace those from ASML in a crucial step in the process of creating the most advanced semiconductors, and do it more cheaply, according to Substrate founder James Proud.
Proud gave some evidence to back up his claims, but said that he was keeping a lot of details to himself for competitive and security reasons. That makes sense.
But to deliver on what he says, and to eventually turn the invention into a commercial enterprise that wins the trust of trillion-dollar companies and their orders, he’ll have to provide much more concrete proof.
For context, Intel Corp. — once the most powerful company in the chip industry and owner of its best factories — said it would reach certain technical milestones and that other chipmakers would flock to give it orders. Those other companies gave it a look and weren’t moved. Reputation counts for nothing.
When he joined Ed on Bloomberg Tech for a TV interview right after unveiling his startup, Proud recognized the path he’s on.
“We have what is a production-quality” chipmaking machine, he said, suggesting Substrate’s technology is at the stage where it’ll be able to to produce in large volumes and in high enough efficiency to start taking orders. “Of course, we are not in fab, we have not built a fab yet, so the proof is really going to be there. But we feel very, very confident on the key metrics that go into throughput — we’ve been demonstrating those.”
Inside some of that technical talk is the assertion that his machine is ready to take its place – or rather ASML’s place – in a production line, but he hasn’t done it yet.
ASML has a well developed customer base, strong order book and global supply chain behind it. Substrate is still a lab experiment.
Chipmaking is expensive because the machines cost so much money and become obsolete so quickly. Producing a tool that works is only step one. Producing a tool that reliably handles an incredibly large volume of work as part of a factory that must churn out millions of chips a month is a step beyond that.
So far, Proud has shown off something that he says proves his machine does a relatively small part of the whole operation. It’s arguably the most difficult part on a basic science level — burning lines that are smaller than the wavelength of visible light — but that doesn’t matter unless everything else that must go with that is in place, too.
From the world of sell-side analysts, Bernstein took the historical view. “In theory, X-ray lithography should offer better resolution than EUV,” analyst David Dai wrote, “but too many problems proved insurmountable.”
Bernstein indicated that becoming a chipmaker as well as equipment maker is probably an ambition too far.
“If the startup truly believes in its technology,” Dai added, “the only viable approach is to form an ecosystem around it and get the industry to work together.”
Substrate’s plan is the opposite — to go it alone. That would indicate that it’ll be able to go longer without facing the (literally) microscopically granular scrutiny of trying to sell its machines in direct competition to ASML. A potential customer such as TSMC would be more than capable of running the rule over any new tool and turning an electron microscope on whatever it produces.
But even then, as a full-fledged chipmaker, Substrate would find itself having to give very detailed answers in an uncompromising world. It would have to produce a huge number of chips that meet incredibly stringent specifications, on time and on budget. In the chip industry, the ability to deliver predictable quality and make rapid improvements is the preserve of a tiny handful of companies that have spent decades and hundreds of billions of dollars perfecting the art. And even then, one bad decision can be disastrous. Just look at Intel’s earnings reports for all the evidence than anyone could need.
SemiAnalysis, an industry research firm that says it has followed Substrate since its early days, deems the timeline that Proud has laid out as “fabulously ambitious” if not unrealistic given industry norms.
Between the believers and the skeptics sits a $100 million bet that venture capital investors have placed on Substrate. A bet that physics, economics and ideology can finally align. To turn that into a working US chip fab will take tens of billions more — and that’s the part no one is debating.
| |
|
|
