Intel’s new weapon, the acquisition of artificial intelligence chip maker Habana Labs

Intel Corporation announced the acquisition of Habana Labs for $2 billion. The company, headquartered in Israel, is a provider of programmable deep learning accelerators for data centers. The acquisition will strengthen Intel’s artificial intelligence portfolio and accelerate its development in the fast-growing emerging market of artificial intelligence chips.Intel expects this market to exceed $25 billion by 2024[i].

Navin Shenoy, executive vice president and general manager of Intel’s Data Platform Group, said: “This acquisition advances our AI strategy to provide customers with solutions for every performance requirement, from the intelligent edge to the data center. Specifically, with a family of high-performance training processors and a standards-based programming environment, Habana Labs has greatly enhanced the capabilities of our data center AI offerings to handle changing AI workloads.”

Intel’s AI strategy is based on the belief that leveraging the power of AI to improve business outcomes requires a broad portfolio of hardware and software technologies and a complete ecosystem to support it. Now, Intel AI solutions are helping customers turn data into business value and generating significant revenue for the company. In 2019, Intel expects AI-driven revenue to exceed $3.5 billion, a year-over-year increase of more than 20%. Together, Intel and Habana will bring best-in-class artificial intelligence products to data center acceleration to meet the evolving needs of customers.

Nayi Sun also said: “We understand that customers looking for easy-to-program, purpose-built AI solutions also require excellent scalability across a variety of workloads and neural network topologies, and Habana excels in these areas. With a proven track record, we are delighted that AI teams like Habana are joining Intel. Our combined intellectual property and expertise will deliver unparalleled computing performance and efficiency for AI workloads in the data center.”

Following the acquisition, Habana will operate as a separate business unit and will continue to be led by the current management team. Habana will report to Intel’s Data Platforms Group, which is also home to Intel’s broad data center AI technology. The acquisition will enable Habana to fully leverage Intel’s AI capabilities, which include deep expertise and substantial resources accumulated over the past three years in AI software, algorithms and research, and will help Habana continue to expand and accelerate its growth.

Habana Chairman Avigdor Willenz agreed to serve as a senior advisor to the business unit as well as Intel. Habana’s headquarters will remain unchanged in Israel, where Intel also has long-term investment and layout. Prior to this transaction, Intel Capital was an investor in Habana.

David Dahan, CEO of Habana, said: “We are fortunate to learn about and work with Intel’s business because of our prior investment, and we are excited to officially join the Intel family. Intel has created a world-class AI team. and capabilities. We are excited to partner with Intel to accelerate and expand our business. Together, we will bring more AI innovations to our customers faster.”

Going forward, Intel plans to leverage its growing AI technology portfolio and AI talent to deliver unparalleled computing performance and efficiency to customers’ AI workloads.

Habana’s AI training and inference products

Habana’s Gaudi AI training processor is currently sampling for select hyperscale customers. Gaudi-based large-node training systems are expected to increase throughput by a factor of 4 compared to systems built with the same number of GPUs. Gaudi aims to make the system achieve efficient and flexible horizontal and vertical expansion.

In addition, Habana’s Goya artificial intelligence inference processor has been commercialized, showing its own excellent inference performance with ultra-high throughput and ultra-low real-time latency in a very competitive envelope power. With the continuous increase of artificial intelligence workloads, the requirements for computing, memory and connectivity are also increasing. Gaudi training products and Goya inference products can provide a rich and easy-to-program development environment, which can help customers deploy differentiated solutions plan.

[i] Intel estimates the total addressable market (TAM) for AI silicon by 2024 will be greater than $25 billion, and within that, AI silicon in the data center is expected to be greater than $10 billion in the same timeframe.

Intel estimates that the AI ​​chip market size (TAM) will exceed $25 billion by 2024, of which the data center AI chip market will exceed $10 billion.

Edit: muyan