In a significant development echoing the intensifying race for artificial intelligence supremacy, Etched, the pioneering AI chip startup co-founded by three Harvard dropouts in 2022, has announced the successful closure of a $300 million Series C funding round. This latest capital infusion propels the company’s valuation to an astonishing $10.3 billion, as confirmed by co-founder and COO Robert Wachen to TechCrunch. The substantial investment underscores the escalating demand for specialized AI hardware and the venture capital community’s profound confidence in Etched’s innovative approach to AI inference acceleration.
The funding round was spearheaded by Sequoia, a prominent venture capital firm known for its early bets on transformative technology companies. Joining Sequoia were an impressive roster of leading investors, including Andreessen Horowitz, a16z, another titan in the VC world; SK Hynix, a global semiconductor powerhouse, signifying strategic industry interest; Jane Street, a quantitative trading firm; and Diffusion Capital. Additionally, several earlier investors reiterated their commitment by participating in this round, further solidifying Etched’s financial backing. The company’s formidable list of early individual backers includes influential figures such as Peter Thiel, co-founder of PayPal and Palantir; Andrej Karpathy, a leading AI researcher previously at OpenAI and Tesla; Dylan Field, CEO of Figma; and Amjad Masad, CEO of Replit, among others. Such a constellation of strategic and financial support highlights the broad belief in Etched’s potential to carve out a significant niche in the burgeoning AI hardware market.
A Rapid Ascent in the AI Chip Landscape
Etched’s valuation trajectory has been nothing short of meteoric. Just seven months prior, in December, the company secured a $500 million funding round, which valued it at $5 billion. The current Series C round, doubling its valuation in such a short span, reflects an extraordinary acceleration in investor confidence and market perception. This rapid appreciation places Etched in an elite category, with the company noting that this marks the highest valuation ever achieved for a Sequoia-led Series C funding round. This milestone not only validates Etched’s technological advancements and business strategy but also signals a broader trend of significant capital flowing into deep tech, particularly within the AI hardware sector.
The announcement of this funding round comes on the heels of several critical operational achievements. Last month, Etched revealed it had successfully manufactured its homegrown chips, a pivotal step from design to tangible product. Furthermore, the company disclosed that its first full systems were undergoing testing by clients, indicating a readiness for real-world deployment. Crucially, Etched had already booked an impressive $1 billion worth of orders, demonstrating strong market demand and early commercial traction even before this latest funding round. These accomplishments collectively paint a picture of a company rapidly transitioning from a promising startup to a formidable player in the competitive AI hardware domain.
The timing of Etched’s inception in 2022 was prescient. It emerged at a juncture when the concept of building chips specifically optimized for AI models based on transformer technology—the foundational architecture underpinning most modern large language models (LLMs) like ChatGPT and Claude—was largely considered an unconventional, if not audacious, endeavor. At the time, the market was heavily dominated by general-purpose GPUs, primarily from Nvidia, which had become the de facto standard for both AI training and inference. Etched’s initial focus on specialized silicon faced skepticism, with some questioning the viability of a hardware solution tailored to specific AI paradigms.
Redefining AI Inference: Etched’s Technological Edge
Despite the initial perception, Robert Wachen clarified that Etched’s products, which are sold as complete systems rather than standalone chips, are designed to be far more versatile than initially understood. The company’s systems are capable of running a broad spectrum of AI models, extending beyond just traditional transformer architectures. This includes sophisticated Mixture of Experts (MoE) models, such as DeepSeek and Qwen, which dynamically split tasks across specialized sub-models for enhanced efficiency and performance. Moreover, Etched’s hardware supports non-transformer designs like Mamba, which is built upon a distinct underlying architecture known as a state-space model, showcasing the adaptability of their specialized hardware to evolving AI paradigms.
Interestingly, the idea of "etching" parts of specific AI models directly into silicon to boost performance, initially seen as radical, is now gaining traction within the industry. Reports indicate that technology giants like Google are reportedly exploring similar concepts with their "Frozen v2" chip, aiming to embed Gemini’s architecture directly into silicon. This development from a major player like Google provides significant validation for Etched’s original, albeit misunderstood, vision and underscores a broader industry shift towards highly specialized, application-aware hardware designs.
Etched’s core innovation lies in its redesign of two critical components from the ground up, specifically to accelerate AI inference—the computational process that executes an AI model after a user submits a prompt, generating responses or predictions. Wachen explained that inference fundamentally operates in two distinct stages: "prefill" and "decode."
The "prefill phase" is computationally intensive, focusing on understanding the input prompt, establishing context, and processing the initial query. This stage involves complex mathematical operations to encode the user’s input into a format the AI model can process. Recognizing its demanding nature, Etched developed a specialized prefill chip designed to operate "dramatically" faster than existing solutions. This speed enhancement is achieved by running the chip at a significantly lower voltage than conventional AI chips, a technique Etched terms "low-voltage inference." The inherent advantage of lower voltage operation is reduced heat generation, which in turn allows for a higher density of transistors to be packed onto the chip, further boosting computational power and efficiency without compromising thermal management.
For the subsequent "decode phase," which is responsible for generating the output tokens—the actual answer or content the user sees—the primary challenge shifts from raw computation to massive memory bandwidth and low-latency access. This stage requires rapid retrieval and manipulation of large amounts of data stored in memory to sequentially generate the output. To address this, Etched engineered a novel type of memory and interconnect technology, which it calls "cluster-scale memory." This innovative architecture enables multiple chips to connect seamlessly and utilize a shared memory pool with exceptionally high speed and ultra-low latency. The combined result of these two specialized components, as promised by Etched, is a significant leap in inference speed coupled with a substantial reduction in operational costs, offering a compelling value proposition for AI service providers.
Strategic Investments and Industry Validation
The participation of leading venture capitalists and strategic investors like Sequoia, Andreessen Horowitz, and SK Hynix is a powerful endorsement of Etched’s technology and market potential. Sequoia’s decision to lead the round, especially at such a high valuation for a Series C, signifies a strong belief in Etched’s ability to disrupt the existing AI hardware landscape. Sonya Huang, a partner at Sequoia, has previously articulated the firm’s thesis around the need for specialized hardware to meet the escalating demands of AI, especially for inference workloads. Her involvement, and that of other partners like Abishek Malani, underscores the strategic alignment between Etched’s vision and Sequoia’s investment strategy.
The inclusion of SK Hynix, a major memory chip manufacturer, is particularly noteworthy. This investment could signal a strategic partnership opportunity, potentially providing Etched with crucial supply chain advantages and deep expertise in memory technology, which is critical for their cluster-scale memory innovation. Such a partnership could accelerate Etched’s path to mass production and market penetration.
Beyond financial backing, Etched has garnered validation from some of the most respected minds in the AI community. Robert Wachen highlighted that individuals like Andrej Karpathy (formerly of Anthropic, OpenAI, and Tesla), Noam Brown (OpenAI), and the legendary "Godfather of AI" Geoffrey Hinton have all had direct experience with Etched’s hardware through private demos. Their excitement and endorsement, stemming from firsthand interaction with the technology, lend significant credibility to Etched’s claims of superior performance. This direct validation from industry luminaries is a powerful testament to the tangible benefits and potential impact of Etched’s specialized AI systems.
From Garage to Global Player: The Founders’ Tenacity
The journey of Etched’s founders—CEO Gavin Uberti, COO Robert Wachen, and Chris Zhu—is a classic Silicon Valley tale of audacious vision, relentless perseverance, and overcoming formidable challenges. The trio famously dropped out of Harvard University to pursue their ambitious startup, embarking on a path fraught with unknowns. Wachen candidly admitted that they "had no idea how hard it was going to be," particularly concerning the complexities of raising substantial capital and building a world-class engineering team from scratch.
The early days were characterized by extreme resourcefulness and personal sacrifice. Wachen recounted his arrival in the Bay Area, having left Harvard, without an office or even an apartment arranged. He vividly described sleeping on the floor of a friend’s unfurnished house, using a towel as a makeshift blanket. The initial server infrastructure required to run their intricate chip-design tools was set up in the garage of an early employee. In a testament to their dedication and the nascent stage of the company, whenever the servers needed a reboot, the employee would call his wife, who would then physically go and press the reboot button. These humble beginnings starkly contrast with the company’s current status and valuation, underscoring the founders’ unwavering commitment.
Today, Etched is a rapidly expanding enterprise, a stark departure from its garage origins. The company now employs 400 people, bustling within a professional office environment. It operates a significant 2-megawatt data center and has recently opened an expansive new 80,000-square-foot, 10-megawatt facility in Milpitas, strategically located near its main San Jose office. This rapid expansion of infrastructure reflects the company’s aggressive scaling plans and its commitment to developing and deploying its cutting-edge AI systems. Wachen proudly stated, "We’re running tokens in our lab today, working with some of the largest AI companies in the world," indicating that Etched is actively engaged with key industry players.
Reflecting on his journey, Wachen light-heartedly quipped about his current living arrangements, "I also have a blanket now, and a mattress, and a pillow even. Multiple pillows." More profoundly, he emphasized the critical lesson learned from their arduous journey: the importance of never letting doubters deter them. "It’s come a long way. It’s a very, very different world. But I think, when you really think something’s possible, and you just work at it for a long time, you can do it," he concluded, offering an inspiring message about the power of conviction and persistent effort in the face of adversity.
Broader Implications for the AI Hardware Ecosystem
Etched’s remarkable success and rapid valuation increase have profound implications for the broader AI hardware ecosystem. For years, Nvidia has held a near-monopoly on high-performance GPUs, essential for AI training and inference, largely due to its CUDA software platform. However, the emergence of highly specialized AI chip companies like Etched signals a potential shift in this paradigm. While Nvidia’s dominance in training may remain unchallenged for some time, the inference market, which represents a massive and growing opportunity as AI models move into widespread deployment, is ripe for disruption by purpose-built hardware.
Etched’s focus on optimizing inference through innovations like low-voltage prefill chips and cluster-scale memory addresses critical bottlenecks in current AI deployments: power consumption, latency, and cost. As AI applications become more pervasive and require real-time responses, the efficiency gains offered by specialized inference chips will become increasingly valuable. This trend suggests a future where data centers might deploy a more heterogeneous mix of hardware, with general-purpose GPUs for training and specialized accelerators like Etched’s systems for inference.
Furthermore, Etched’s journey validates the venture capital market’s appetite for "deep tech" investments, particularly in areas that promise significant architectural innovations rather than incremental improvements. The willingness of top-tier VCs to pour hundreds of millions into a hardware startup at such an early stage and at such a high valuation underscores the strategic importance of controlling the underlying compute infrastructure for AI. It also highlights the recognition that software innovation alone cannot sustain the exponential growth of AI without corresponding breakthroughs in hardware.
The company’s story serves as an inspiration for other ambitious startups daring to challenge established norms and tackle complex engineering problems. In a technology landscape often dominated by software and platform plays, Etched’s hardware-centric approach demonstrates that foundational innovation in silicon remains a powerful avenue for creating immense value and impact. As AI continues to evolve, companies like Etched are poised to play a crucial role in shaping its future, driving efficiency, performance, and accessibility across a myriad of applications.
About the Author
