Join executives from July 26-28 for Transform’s AI & Edge Week. Hear from top leaders discuss topics surrounding AL/ML technology, conversational AI, IVA, NLP, Edge, and more. Reserve your free pass now!
When Credo Technology went public earlier in mid-February, raising $230 million at a value of $1.6 billion, little did we know that it was going to be such a rare event for 2022.
Initial public offerings (IPOs) have since dried up in current stock market malaise, and IPOs for semiconductor companies are a rare breed anyway. Credo has cash and it is just becoming profitable as well. And it owes this to the growth of the data economy.
Credo makes chips and hardware solutions that provide high-speed connectivity for electrical and optical Ethernet applications in the 100 gigabit to 800 gigabit markets.
That means the datacenter. Credo’s custom semiconductor products use digital signal processors (DSPs), and SerDes (serializer/deserializer) chiplets, or components of chips. They also use active electrical cables and other chips.
Since there is a huge amount of ongoing data creation, analysis, and consumption, Credo believes the data economy will continue to expand on an exponential curve on that front, said Bill Brennan, CEO of Credo, in an interview with VentureBeat. But that expansion has to be seen against the backdrop of a weakening global economy and growing inflation.
“At a macro level, I’m well aware of the macro challenges that we’re facing as we come off of the post- pandemic reality, as well as what’s happening in China that has really slowed down the consumer end of that business,” Brennan said.
Attacking an opportunity in connectivity
The data infrastructure market where Credo plays is driven by 5G infrastructure, high-performance computing, and hyperscale datacenters (like Microsoft, Google and Amazon Web Services). In this market, the whole ballgame is about balancing power efficiency and cost.
Credo’s customers are in HPC, 5G wireless networking, cloud infrastructure, and hyperscalers. The customers include computer makers, optical module manufacturers, and hyperscalers. Over 14 years, Credo has grown to more than 400 employees.
Back in June, long after the February IPO, Nasdaq invited the company management back to ring the opening bell in person, as it wasn’t possible to do it earlier due to COVID-19.
“It was an overdue celebration for the IPO,” Brennan said.
The company’s 2022 revenue was $106.5 million, up 80% over $58.7 million a year earlier. And FYQ4 revenues were $37.5 million, up 90% from a year ago for the fourth fiscal quarter ended April 30. The quarterly GAAP net loss per share was 4 cents, or $5.4 million, while non-GAAP net income was two cents a share, or $2.8 million in profit. The company took 14 years to hit the $100 million revenue mark, and it expects to hit $200 million in one more year.
From unicorn to IPO
Brennan noted that people said the company has become a unicorn (with a valuation above $1 billion), but he thinks of the company more as a workhorse, with a different kind of personality than most unicorns, which tend to trade more on future promise than current reality.
“Our trajectory as a company might be a little bit different than the trajectory of the classic Silicon Valley startup,” he said. “If you have seen our last results, we grew at a rate of more than 90% on revenue, and we established profitability a lot earlier than we expected.”
Brennan said the company’s ability to delivery energy efficiency and disruption in the market is key.
“As we look forward to our next fiscal year, we expect to grow at that same rate but really grow our profitability even faster into the double-digit rate,” Brennan said. “We find ourselves among the fastest growing semiconductor companies in the market. And we’ve established scale to the point where we can start realizing profitability. So I think we’re at the point where, as we face this upcoming challenge, I think we’re at the point where we find ourselves on the leading edge of the market.”
Rivals include Marvell, Allegro Microsystems, Silicon Laboratories, Alphawave, Tower Semiconductor, Synaptics, Broadcom, and more. Credo is fabless, meaning it designs chips and uses contract manufacturers to build them. That reduces its margins, but it also protects it from supply and demand swings that can affect the profitability of factories.
The company went public in February at $10 a share and its stock rose as high as $16.90 a share by the end of February, but since that time it started a steady decline. Now the stock has bounced back some and it is floating just above the IPO price. The lockup from the six months after the IPO is
set to expire shortly, and employees and management will be able to sell as many as 100 million shares shortly, well above the 20 million issued in the IPO.
The San Jose, California-based company was founded in 2008 by veteran engineering managers Lawrence Cheng (Credo’s chief technology officer) and Job Lab (chief operating officer). Lip-Bu Tan, former CEO of Cadence Design Systems and founder of the Walden International investment firm, started working with the founders in Shanghai as an investor about six years ago, and Brennan became president of the company in 2014. That turned out to be a good time for the company.
Brennan joined Marvell in 2000 and managed its business of storage-related chips for 11 years. He noted that the company lost three key engineering directors in 2008. And two of them had been critical at Marvel, and they turned out to be the founders of Credo. Tan became chairman in January 2022.
“I’m very proud to be associated with Credo,” Tan said. “I joined the board and became the chairman of the board. Credo is basically addressing high-speed connectivity for the datacenter cloud. High-speed connectivity is a must.”
Tan also noted how the company became a leader in chiplets, or subcomponents in a chip or package of chips. Tan noted you can add many millions of transistors more using the chiplet approach where you can provide hig-speed connectivity between the components.
“All the hyperscale companies are building up massive stacking of data, switches, storage, and infrastructure,” Tan said. “It’s more elegant to try to improve the throughput,” or lifting the speed limit on a freeway as opposed to adding more lanes.
“We’re the onramp and the offramp for the highway,” Brennan said.
Tan said the company had a more advanced team with the ability to design a device with a better balance between the various tradeoffs.
“I call it an art, not a science,” he said. “You can’t match their design sophistication. This is really an A team. The secret part is basically to be the ultra-low power and high speed.”
Credo focuses on transmitting data over a wire and receiving it on the other end. The wire can be copper wire or it can be optical. That means taking data from the digital realm, transmitting it in analog form over the wire, and then moving it back to the digital domain at the end of the wire.
“The challenge is in the transmit and receive,” Brennan said. “Any wire has loss associated with it. As it loses signal, you find yourself with a huge signal processing challenge. It’s a huge problem we are solving. The other key is doing this with low power.”
The datacenter explosion
In the past five years or so, datacenters drove the need for exponential increases in networking speed and computing. The networking speeds moved beyond 10 gigabits a second, where they had been stuck for some time, to 25 gigabits a second in 2016 and 100 gigabits a second now. Now engineers are working on 200 gigabits a second.
“The datacenter drove the need for exponential speed increases,” Brennan said. “It’s this exponential curve that makes the problem really hard. It’s been a long and winding road, and we found ourselves in a really unique position, not only to address the datacenter, but the rest of the data infrastructure market as well.”
Eventually, Brennan believes this revolution in networking speeds at the datacenter will also make its way down to consumers as well. The company will keep driving on next-generation interfaces to improve networking speeds. The company will focus on architectural improvements that can keep up with the demands in the market.
“That’s really the core of why we’re the disrupter and the pioneering in connectivity,” Brennan said.
The future of chips
As for the chip industry, Tan and Brennan are upbeat.
“Semiconductors have become very strategic for the country and very important for the whole world,” Tan said. “President Biden has brought up chip wafers and this chip act is very important for the country and the world.”
Besides chips, Tan is paying a lot of attention to AI and machine learning. He believes it can be used to improve the world in everything from drug discovery to personalized medicine and improvements in all of the verticals like banking and financial services.
“I call this the data-centric era,” he said. The connectivity of data is very important, as is the transmission and the latency.”
He noted the critical energy-related problems of climate change and the war in Ukraine, which has driven up energy prices.
He the current market is challenging with inflation growing and the choppy stock market. He said he was grateful that Credo has been trading near the IPO price. He thinks chips are critical in terms of their value to the economy.
“I believe this is the foundation we need to have in terms of scaling other applications,” he said. “I invest in high-speed connectivity. You need to have a full stack. You have to have the software, the compiler, the algorithm, and all of it to make it really effective. This is the frontier of semiconductors.”
Tan said he believes that the big companies sometimes move too slow and that startups can move faster.
The chip cycle and supply chain issues
As for the chip industry cycle, he noted that the bull run in semiconductors has gone on for a long time. Now that smartphones have hit a huge installed base, sales may slow down and there could be softness in the end market for chips. He said he can see the market slowing down in 2022 or 2023. But he believes that it could bounce back by 2024.
“I think the next two years will be challenging,” Tan said, thanks to issues such as inflation and supply-chain issues. “The other part is the geopolitical risk. This relates to energy and political tensions.”
And in the long term, he is concerned that progress in chip design could also slow down as the chip industry hits the limits of Moore’s Law, the prediction that chip capacity would double every couple of years. Good relationships with partners in electronic design automation and foundries have become critical. Credo is a close partner for contract chip manufacturer TSMC.
“It is more tricky to add performance,” he said. “With respect to Credo, there is no need to push the advanced manufacturing technology. They have a huge advantage in design. With Moore’s Law slowing down, you need to do chiplets and clever packaging and that tis why Credo saw the chiplet as the right way to go.”
As for the explosion of data, and whether tech can keep up with it, Brennan said, “We are living in a world right now where we are faced with endless opportunity. The compute world is at the knee in the curve, and I don’t see that stopping. I think we’re all going to be extremely happy in the next 10 years.”