Internet-enabled consumer apps may have hogged the limelight, but the next decade could belong to the semiconductor industry, said Pradeep Vajram, chairman and chief executive at AlphaICs, a California- and Bengaluru-based chip maker. AlphaICs recently raised $8 million in a series B funding to develop a portfolio of chipsets for artificial intelligence workloads.
While behemoths like Qualcomm, Intel and Samsung dominate the industry, recent times have seen a spurt in startups producing niche chips for heavy-duty computing like machine learning and big data analytics. The new generation of chips are designed to support parallel processing at low power and latency, making them attuned for driverless cars, images and language processing and high-end AI tasks at scale.
India’s engineering talent and ability to create intellectual property (IP) is more sophisticated than ever, but the growth on the hardware side for India-made products, including chips, has been tepid. “Silicon is a deep market, you need at least $20 million, and a long gestation period, to develop prototypes. VCs here don’t have that appetite,” said Vajram. Most of the talent is employed locally at large MNCs, which is also a good thing, he added.
AlphaICs has one another ace up its sleeve--Vinod Dham, the ‘father of Pentium chip’ is one of its co-founders. He has been key driver in creating the blue-print of AlphaICs product.
“In the early 2000s, many MNCs (multinational corporations) came to India for cost reasons and set up captive centres. Twenty years on, these centres have evolved into full product units, ideating and shipping new products to the world,” said Vajram. To spurt chip startups from India, a success story like Flipkart in e-commerce or Infosys in Information Technology could be a catalyst. “We only need one,” said Vajram.
India’s semiconductor market is currently estimated at $53 billion (2020), according to Indian Semiconductor Association, part of the $422 billion global market size, as per IDC estimates, 2020. The biggest producer is the US, followed by China.
The early rise in chipsets happened on the back of smartphone proliferation, and now the growth is driven by AI and Cloud computing, which require tailor-made chipsets. “AI is new technology. Traditional CPUs and GPUs can perform parallel processing but they are not very efficient in cost and power consumption. Therefore they are new architectures coming up,” said Vajram.
Chip-makers are organised based on the category of products they supply for: there is high-end computing such as cloud and data centres which require high-power chips; there is a low-power segment which has smartphones and hand-helds, and finally there is a fast emerging mid-power segment which includes surveillance, automotive, IoT, Edge servers applications.
Vajram was among the initial investors in AlphaICs, and later joined full-time in 2019.
AlphaICs, which operates in the mid-power segment, estimated to be $15-20 billion, according to Vajram, is designing a chip targeted for vision applications, like surveillance, industrial automation, and IoT. There are two concepts at play here: AI-focused chips and Edge computing. The latter refers to a system design where processing and data storage is closer to the location where it is needed in order to reduce bandwidth use and improve response times.
To that end, AlphaICs has developed what it calls Real AI Processor (RAP), a proprietary architecture that enables AI acceleration for low power edge applications, as well as high performance edge data centres. A characteristic of Edge is that the chips are self-learning which make processing more efficient with time. According to AlphaICs, quoting Omdia, AI edge chipset revenue will grow from $7.7 billion in 2019 to $51.9 Billion by 2025.
“Edge AI applications in consumer markets like high-end smartphones, wearables as well as enterprise markets like robots, cameras, and sensors will be pervasive in the next few years,” said Sateesh Andra, Managing Director with Endiya Partners, an investor in AlphaICs.