Fractile Raises $720 Million Series C at $7 Billion Valuation to Develop Memory-First AI Chips

London-based semiconductor startup Fractile has raised $720 million in Series C funding at a $7 billion valuation to develop memory-first AI chips designed to accelerate artificial intelligence workloads.

The funding round was led by Thrive Capital, Founders Fund, Lightspeed and Redpoint. Other participating investors include Alphabet's CapitalG, DST Global, BlackRock, Bain Capital Ventures, T. Rowe Price, Shopify Ventures, Balderton, the UK's Sovereign AI Fund and the British Business Bank.

According to the Financial Times, which cited people familiar with the deal, the latest funding values Fractile at $7 billion, a substantial increase from its approximately $1 billion valuation in May 2026.

Founded by Walter Goodwin, the company is developing a new chip architecture that combines memory and processing components in a three-dimensional stack. Its objective is to address a major limitation in modern AI hardware: the speed at which processors can access data.

Fractile Founder Walter Goodwin and the Memory Bottleneck in AI

Fractile was founded by Walter Goodwin, who leads the company's development of memory-first computing architecture.

Goodwin's central argument is that improvements in processor performance have significantly outpaced advances in memory bandwidth.

According to the figures cited in the report, chip performance has increased approximately one millionfold over 25 years, while memory bandwidth has grown only around 40 times.

This creates a substantial gap between how quickly processors can perform calculations and how quickly they can access the data required to complete them.

The problem becomes particularly important for large AI models, which need to move significant volumes of data during inference.

Goodwin describes existing AI hardware as being constrained by memory access, arguing that processors are effectively waiting for data rather than operating at their full potential.

How Fractile's Memory-First AI Chips Work

Fractile is developing a chip architecture that combines custom dynamic random-access memory (DRAM) and processing logic in a three-dimensional stack.

The design aims to increase the bandwidth available between memory and processing components while retaining sufficient memory capacity to support large AI models and long context windows.

This approach differs from architectures that rely heavily on high-bandwidth memory (HBM).

HBM is widely used in advanced AI computing systems to move data quickly between memory and processors. However, its cost and supply constraints have become important considerations as demand for AI infrastructure increases.

Fractile believes its architecture could offer a different balance of memory capacity, bandwidth and performance.

The company claims its technology could make AI inference up to 10 times faster. However, this is a company performance claim, not an independently demonstrated production benchmark.

$720 Million Funding to Accelerate Chip Development

The $720 million Series C round brings together major venture capital firms, financial institutions and strategic investors.

Thrive Capital, Founders Fund, Lightspeed and Redpoint led the investment.

Other participants include:

• CapitalG

• DST Global

• BlackRock

• Bain Capital Ventures

• T. Rowe Price

• Shopify Ventures

• Balderton

• UK Sovereign AI Fund

• British Business Bank

The funding follows Fractile's approximately $220 million raise in May 2026, when the company was valued at roughly $1 billion.

The latest reported valuation of $7 billion represents a substantial increase within a relatively short period.

The investment reflects growing interest in alternative AI chip architectures as demand for computing infrastructure expands.

Anthropic Interest and the AI Inference Opportunity

The Financial Times reported that Anthropic has expressed interest in becoming an early customer for Fractile's chips, citing people familiar with the discussions.

This potential interest is significant because AI model developers need increasingly efficient infrastructure to serve large numbers of users.

Inference involves generating responses when users interact with AI models. As AI adoption expands, inference workloads are becoming a major part of the computing infrastructure market.

Fractile is particularly targeting workloads involving AI agents, which may operate continuously for extended periods and require repeated model interactions.

Thrive Capital's Philip Clark sees the inference market as a major opportunity, potentially becoming at least as large as the market Nvidia currently addresses.

However, Anthropic's reported interest should not be treated as a confirmed commercial agreement or customer contract.

Fractile Faces Competition in the AI Chip Market

Fractile is entering an increasingly competitive semiconductor market.

Nvidia remains a major player in AI computing, while companies such as Etched, MatX and British startup Olix are developing alternative chip architectures.

The market is also attracting attention from companies seeking to reduce dependence on existing GPU and memory supply chains.

Fractile's differentiation lies in its memory-first architecture and its focus on reducing the data-transfer bottleneck that can limit AI inference performance.

If the company can demonstrate its claimed performance improvements, its technology could become relevant for AI infrastructure providers looking for faster and potentially more efficient inference systems.

Building the UK's AI Hardware Ecosystem

Fractile's funding round is also significant for the UK's ambition to develop domestic AI infrastructure.

The company is headquartered in London and also operates offices in the United States and Taiwan.

The participation of the UK's Sovereign AI Fund and British Business Bank reflects institutional investment in developing advanced AI hardware capabilities.

However, Fractile is building for a global semiconductor market, where competition will depend on manufacturing execution, chip performance, reliability and commercial adoption.

What Fractile's $720 Million Funding Means for AI Infrastructure

Fractile's Series C funding highlights investor interest in solving the infrastructure bottlenecks emerging as AI models become larger and more widely deployed.

Its memory-first chip architecture is designed to address the growing gap between processor performance and memory bandwidth.

The company now has substantial funding to advance its technology towards commercial production. Its first chip is planned for 2027, but the claimed 10-fold inference improvement has yet to be demonstrated in a production product.

The next major milestone will be turning its prototypes into working chips that can deliver the promised performance at commercial scale.

For Fractile, the opportunity is not simply to build another AI processor, but to change how memory and computing are designed together for the next generation of AI infrastructure.