No product company, why 16.7 billion in half a year?

Author: Leman

Editor: Wang Qingwu

Manual input丨88%   AI content丨12%

In mid-October, Reuters reported, citing two insiders, that Nuvacore, a chip startup established for only six months, is raising about $200 million at an estimated valuation of approximately $2.5 billion, which is about 16.7 billion yuan, for developing a new type of central processing unit aimed at AI data centers. This round has not yet closed, and the valuation and amount may still change.

It is worth noting that Nuvacore currently has no products, no revenue, and no performance testing reports, having previously completed a seed round led by Sequoia Capital. Now financing with a unicorn valuation inevitably draws attention to the people behind it.

The core figures of Nuvacore are mainly three: Gerard Williams, John Bruno, and Ram Srinivasan. These three formed a team that has built CPU cores at Arm, Apple, and Qualcomm for over twenty years, with Williams having the strongest credentials: during his time at Apple as the chief CPU architect, he worked on multiple generations of self-developed chips from A7 to M1.

Friends familiar with Apple might know the importance of these two generations of chips to Apple. The A7 was Apple's first 64-bit mobile chip and also marked the smartphone industry's first large-scale shift to a 64-bit architecture, allowing Apple to leapfrog its competitors in performance by an entire generation; the M1 was the first chip for Apple’s Mac product line to transition from Intel to self-developed chips, which significantly changed Mac’s shipment structure, gross margin levels, and chip power dynamics over the next four years. These two chips defined Apple's architectural shifts in the mobile and PC segments, with Williams involved in both.

The third startup journey of three chip veterans

The founding team of Nuvacore has a story that starts with Arm.

Williams worked at Arm for 12 years and was the technical lead for Cortex-A8, Arm's first superscalar core, which was later used in the iPhone 3GS. He also participated in crucial architectures such as Cortex-A15. Around 2010, he switched to Apple, serving as a senior director of platform architecture for nearly a decade, responsible for CPU and SoC development, and is the inventor of over 60 Apple patents.

In February 2019, Williams left Apple and co-founded Nuvia with John Bruno and others, aiming to create high-performance Arm CPU cores for servers.

This startup lasted only two years, but the result was not bad: the company raised a total of $293 million and was acquired by Qualcomm for $1.4 billion in cash in early 2021, with its design later becoming the foundation for Qualcomm's Oryon architecture. Williams joined Qualcomm with the deal, becoming senior vice president of engineering.

It was also during the Nuvia period that the team first tasted the challenges of architectural choices. In 2022, Arm sued Qualcomm over how to use Nuvia permissions after the acquisition. There were multiple court hearings with mixed outcomes, and a unified ruling on the Nuvia licensing issue has yet to be reached, requiring further court handling.

The turning point occurred at the beginning of 2026, when Williams announced his departure from Qualcomm. Soon, he and his old partners—John Bruno (whose career began at ATI, later joined AMD after ATI was acquired, spent five years in Apple’s platform architecture team, and then worked at Google as a system architect) and Ram Srinivasan (who was a performance architect at Intel for seven years, later responsible for system performance architecture at Apple, then joined Nuvia, and entered Qualcomm via acquisition)—reassembled to establish Nuvacore.

This time, their slogan was more impressive, claiming they would “rewrite the rules of silicon chips.” The team quickly expanded. From April to September this year, the company successively appointed senior vice presidents in software, operations, and marketing, with former Apple CPU engineering head David Williamson taking on the role of senior vice president of hardware engineering.

On September 29, the company announced its technical roadmap on social media: the first core design is named WarpCore and adopts a “Core First” strategy opposite to traditional chip development processes—designing core IP such as execution units, branch prediction, cache, and memory subsystems first, and then deciding on using x86, Arm, or RISC-V instruction sets. The company stated that day, “We said we would rewrite the rules of silicon chips. We mean it.”

This strategy is quite disruptive. But it has its technical basis: modern high-performance CPUs translate software instructions into internal micro-operations for execution, with the instruction decoder being the only module in the chip tied to the instruction set. Jim Keller, a well-known figure in chip design, mentioned in a speech in 2022: “All modern computers are actually RISC machines; the only thing you need to change is the decoder.” Nuvacore claims that over 95% of the core IP can be developed before selecting the instruction set.

However, this line of thought has had some not very successful precedents in the industry. Around 2014, AMD simultaneously advanced x86 and Arm lines, namely the SkyBridge platform and K12 core, but later halted both due to resource dispersion and turned to Zen. However, Nuvacore's approach is the opposite of AMD: not pursuing two instruction sets at the same time, but rather choosing none initially, waiting for back-end maturity before selecting.

Which route is correct currently has no answer. But the route dispute ultimately depends on who is executing it. For a team that has delivered on Nuvia's promise, the market at least is willing to extend some credit.

In six months, the value nearly doubled

Nuvacore’s first financing occurred in April this year, led by Sequoia Capital, and the amount was not disclosed; the second round is the one currently underway: seeking over $200 million at an estimated valuation of around $2.5 billion.

In contrast, the last project of the same team, Nuvia, took about 19 months to reach a Series B round, with the endpoint price being Qualcomm’s acquisition for $1.4 billion in cash. However, just six months after Nuvacore's seed round, the valuation discussions had already reached $2.5 billion, equivalent to 1.8 times the exit price over the entire entrepreneurial cycle last time. Furthermore, Nuvia’s valuation was based on an existing product “Phoenix” that was already in benchmark testing, while Nuvacore has none.

Why are the same three people’s second startup nearly twice as valuable as the first? This has to be viewed from the results of their last startup.

In fact, Nuvia’s product was also never officially released. At the time of acquisition, its Phoenix core was only in the GeekBench beta testing stage, while the server SoC “Orion” was still in development. But the $1.4 billion pricing was proven to be a good deal afterward, as this design was commercialized within Qualcomm, resulting in today’s Oryon CPU.

Looking back at Qualcomm’s official financial report, the Oryon results fundamentally supported its three product lines: PC, automotive, and flagship mobile chips. In the first quarter of fiscal year 2026, Qualcomm recorded a record single-season revenue of $12.3 billion, of which automotive business single-season revenue was $1.1 billion, a year-on-year increase of 15%, continuously breaking records; IoT business revenue was $1.69 billion, a year-on-year increase of 9%; during that quarter, Qualcomm announced plans to launch approximately 150 new models equipped with Snapdragon X series PC processors in collaboration with partners, with the new generation product X2 Elite Extreme becoming the fastest Snapdragon laptop platform to date.

By the third fiscal quarter, Qualcomm's automotive business had maintained double-digit year-on-year growth for 23 consecutive quarters, with combined automotive and IoT revenue growing 28% year-on-year; the CEO provided a target in the financial report, expecting non-mobile business revenue to reach $40 billion by fiscal year 2029, with non-mobile business growth including data centers projected to accelerate from 24% in fiscal year 2026 to over 60% in fiscal year 2027.

It can be said that Oryon's results fundamentally changed Qualcomm's product structure and CPU architecture. Therefore, Nuvacore's $2.5 billion valuation is indicative of this team's potential to once again achieve what Nuvia did in server CPUs.

The market is not lacking in “$2.5 billion” stories

In chip financing, there is a distinctly different market rhythm between domestic and international scenes.

The domestic primary market has begun to retreat from high-profile chip projects. Data from the China Venture Capital Institute shows that the total semiconductor financing in the first half of 2026 dropped approximately 41% year-on-year, with the number of transactions falling by 35% year-on-year, and the difficulty of raising large funding rounds in mid to later stages has significantly increased, resulting in longer project financing cycles and a shift in bargaining power from sellers to buyers.

Another set of data from CVSource indicates that by the end of July 2026, there were a total of 912 financing cases in the semiconductor sector, targeting 823 companies, with an amount of 162.7 billion yuan. While the absolute scale is not low, the structural characteristic of “funds concentrating on leading and specific segments” is evident.

In contrast, overseas conditions are the opposite: global semiconductor startup financing continues to operate at a high level, with both the density of large rounds and the surge in valuations of leading projects being much more aggressive compared to domestic markets.

Specifically, for the leading projects in the AI chip sector, pricing has entered a new order of magnitude. The most recent comparison comes from Tenstorrent, a RISC-V chip company led by Jim Keller, which announced on its official website on October 19, 2026, that it completed a $1.5 billion Series E financing at a valuation of $6.2 billion.

Etched went even further: it previously completed financing at a valuation of $21 billion, and according to TechCrunch, the investment offers it has recently received have reached a valuation of $40 billion or more.

Another company more comparable is Rivos: also founded by former Apple chip teams, also targeting data centers, and also involved in a lawsuit with Apple: in 2022, Apple sued Rivos for stealing trade secrets, and the two sides reached a settlement in February 2024. Rivos completed a $250 million Series A round in April 2024 (led by Matrix Capital, with participation from Intel Capital, MediaTek, Dell Technologies Capital, BlackRock, among others). In October 2025, Meta announced the acquisition of Rivos at a valuation of about $2 billion.

When looking at these transactions alongside Nuvacore, a $2.5 billion valuation does not seem out of place in the overseas semiconductor startup pricing spectrum.

However, the high-level operation overseas is largely driven by strategic investments from semiconductor giants. According to Crunchbase data, in the first five months of 2026, global semiconductor startups raised about $10.7 billion, approaching the entire annual amount of $12.2 billion in 2025. A more intuitive reflection is the density of large transactions: semiconductor giants participated in over 60 rounds of startup financing over $100 million this year, with 16 of these valuations reaching $1 billion or more, with the total scale of startup financing involving semiconductor companies exceeding $250 billion, several times the previous peak.

So, to be more precise, the overseas financing narrative is not simply “financial VCs falling in love with high-profile chip projects,” but industrial capital investing instead of self-research in the AI computing power competition, using capital to lock in the supply chain. The aforementioned Rivos and Tenstorrent both have these industrial capital or strategic buyers behind their financing.

A decade ago, Silicon Valley VCs had much less interest in hardware than software, citing long development cycles, large capital requirements, and cyclical demand as reasons. Today, AI computing power has pushed hardware back to the table, and the patience for “no products first given valuation” has also been repriced.

Original content from Investment Times

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到[email protected]，本平台相关工作人员将会进行核查。