The Great AI Chip Scramble
Hardware startups have been chasing a simple promise: build a faster, more power-efficient processor for the burgeoning demands of artificial intelligence. Traditional CPUs struggle with the rapid-fire matrix math at the heart of deep learning. GPUs, especially Nvidia's CUDA-tuned cards, have become the de facto standard—but locking developers into a single platform creates its own risks. Investors, though, see a different narrative. As Matthew Lynley reported in TechCrunch, "seven startups gunning for similar areas of this space, many of which have raised tens of millions of dollars, with at least one startup's valuation creeping near $900 million." The list reads like a who's who of AI hardware: Cerebras, Graphcore, Cambricon, Horizon Robotics, Groq, Mythic, and now ThinkForce. Each is betting that a different architectural twist—inference optimization, memory hierarchy, or model parallelism—will give them enough leverage to pry developers away from Cuda's grip.
Alibaba's backing of Cambricon exemplifies the strategic interest. The Chinese e-commerce titan poured financing into the startup, which was reportedly already valued at $1 billion. Over in the United States, Intel Capital matched that ambition with a $100 million stake in Horizon Robotics, a Beijing-based player focused on autonomous driving and edge AI. Groq, founded by ex-Google engineers, raised around $10 million from Social+Capital, a check that seems modest beside the $50 million Graphcore secured in November 2017 led by Sequoia Capital (followed by a $30 million round in July led by Atomico). Mythic, another contender, brought in $9.3 million. The sheer volume of capital flowing into unproven silicon speaks to how deeply investors have internalized the AI thesis. As Lynley observed, "hardware startups face many more challenges than ones built on the back of software." Yet the money keeps coming.
The incumbents are not sitting idle. Google unveiled its next-generation TPU in May 2017, tailor-made for inference and machine training. Apple, never one to leave silicon to chance, designed its own GPU for the next iPhone, promising tighter integration between hardware and on-device AI features. Intel, having acquired Nervana for a reported $350 million the prior August, promised its own Neural Network Processor by year's end. The message from the giants is clear: they, too, are preparing custom silicon to protect their cloud and mobile moats. The result is a market in flux, where each new announcement reshapes the competitive landscape.
ThinkForce's $68 Million Statement
While the Western press focused on Cerebras and Graphcore, a quieter but equally significant milestone was unfolding in China. ThinkForce, a China-based artificial intelligence semiconductor producer, revealed it had closed a series A round worth RMB450 million—approximately $68 million. The round, reported by Global Venturing on December 15, 2017, brought together a syndicate of strategically aligned investors. Yitu Technology, a machine vision developer, led the strategic backing. Sequoia Capital China and Hillhouse Capital, two of the country's most prolific venture firms, joined Yunfeng Capital, the private equity outfit founded by former Alibaba CFO Winston Wen.
ThinkForce's stated ambition is bold. The company claims its technology will accelerate neural network models and deliver a fivefold power and cost saving over rival products. If true, that kind of efficiency gain could be transformative for data centers and edge devices alike. The catch? ThinkForce had not yet launched its chip at the time of the announcement. The company is working on semiconductors that integrate AI algorithms and a dedicated AI hardware platform, but product availability remained future-tensed. Beyond the silicon itself, ThinkForce is also developing chip virtualization technology that would allow a single CPU to run calculations as if it were a network of several computers—a potential stopgap for organizations not ready to commit to custom hardware.
The human stakes are embedded in the financing story, too. Jiao Huiru, director of strategic development at Yitu, offered a rare quote: "Chip technology and talents are China's strategic priority and we want to have a world-class chip team." She framed the investment in broader terms, noting that "in the era of AI, China and the US will be the globe's dual engines." Her commentary underscores the geopolitical dimension of the AI chip race. China has identified semiconductor capability as a cornerstone of its long-term economic and military strategy, and venture money is one of the primary tools for building domestic capacity. Yunfeng Capital's participation, in particular, signals that even former Alibaba leadership sees the strategic value in backing homegrown chip talent.
What This Moment Means
The AI chip landscape of late 2017 was already crowded, but ThinkForce's entry added a new variable. The startup's focus on integrating AI algorithms directly into silicon—and its claim of fivefold efficiency gains—suggests a different engineering philosophy than the more general-purpose approaches of Cerebras or Graphcore. Whether that translates into shipping products and market traction remains to be seen. History is littered with well-funded chip startups that never made it past the prototype phase. Yet the capital amassed thus far—hundreds of millions across a half-dozen contenders—creates a critical mass of talent, engineering effort, and investor patience.
What's striking is how quickly the field has moved. In just a few years, AI had gone from a academic niche to the single biggest driver of venture capital deployment. The TechCrunch article captures the moment's texture: "all eyes may have been on Nvidia this year as its stock exploded higher thanks to an enormous amount of demand across all fronts: gaming, an increased interest in data centers, and its major potential applications in AI." But the real story, as Lynley framed it, was the "subtler" wave of startups each chasing a slice of that AI compute pie. By December 2017, the pie had already been sliced seven ways, with ThinkForce's $68 million round serving as a distinctively Chinese counterpoint to the Western influx.
The question now is whether any of these challengers can unseat Nvidia's entrenched position. The incumbent enjoys two moats: a massive installed base and a developer ecosystem built around Cuda. Breaking that lock-in is "going to be doubly true for startups that are trying to press their hardware into the wild and get developers on board," Lynley noted. Still, the money keeps flowing. Google's TPU, Apple's custom GPU, Intel's Nervana—each incumbents' response signals that the status quo is not beyond challenge. For the startups, the path forward will require not just better chips, but a compelling developer story, early partnerships, and enough financial runway to survive the long road from silicon to volume production.
The Road Ahead
If 2017 was the year the AI chip explosion arrived, 2018 promised to be the year the field began to sorted out. Some startups would run out of cash before reaching production. Others might find niche footholds in edge computing, autonomous vehicles, or specialized inference workloads. ThinkForce's virtualization technology and claimed power savings, if delivered, could carve a niche among cost-conscious data centers. Meanwhile, the incumbents' own custom silicon efforts threaten to erode the very market share the startups are targeting.
What feels certain is that the venture dollars will keep flowing. The AI compute demand shows no sign of slowing, and each new funding round extends the talent pool and the experimental designs. For now, the landscape remains in flux, a crowded field of well-capitalized hopefuls each betting that their particular architectural twist will be the one to unseat the incumbent. The next twelve months will likely reveal which of those bets pay off—and which will join the graveyard of ambitious silicon startups. Until then, the AI chip wars will continue to be fought in boardrooms, fab lines, and the pages of tech journalism, with each new round of funding adding another name to the list of those willing to bet the company on a faster, smarter processor.
Related Articles
- Infinity Raises $15M to Let AI Write the Chip Code Nvidia Built Over Decades
- How Etched Built a $1 Billion AI Chip Business—Before Anyone Knew It Existed
