Morphing Machines, an Indian fabless semiconductor IP company, building reconfigurable manycore processors for new-age applications in Data Centres, AI-ML at the edge,5G/6G telecom etc., has raised $2.76M in a seed round led by Speciale Invest.

The old English adage goes ‘Jack of all trades, master of none.’

In the world of semiconductors, the jack of all trades is an FPGA (Field Programmable Gate Array) - a semiconductor device that can be configured multiple times to do different tasks, but is not excellent in any specific task.

A ‘master’ for a particular task is an ASIC(Application Specific Integrated Circuit), a chip that is really good at one task, but has no flexibility.

What if you could have a chip that is a jack of all trades AND master of almost everything. Morphing Machines’ manycore processor, REDEFINE^TM aims to be just that.

The tides are turning

Historically, companies have adopted a hardware-first approach - creating customized hardware tailored to specific industry verticals. This involves designing and manufacturing chips which are domain specific, each catering to the needs of verticals such as automotive, healthcare, or telecommunications.

However, semiconductor design and fabrication is a long process, taking 3-4 years, costing 100s of millions of dollars each time. And once the chip is taped-out, one needs to ensure that the product revenues must be at least 10X the development cost to justify the high cost and timelines involved. Further, with fixed functions designed for specific tasks within an industry, there is limited flexibility for updates or changes once the hardware is deployed.

Take the example of Apple. Why can’t Apple simply unlock the desired performance and power consumption levels of the iPhone 15 in the same hardware chip that drives the iPhone 11? Instead, Apple has to spend a few hundred million dollars and 18-24 months to make a new chip for every version of the iPhone and then get tens of millions of the chips manufactured. What if Apple can create a few hundred million ‘same’ chips and install them in all of their iPhones and unlock new features over just the software. This would save them enormous amounts of time and money.

At a more technical level, with new applications coming up, the industry began moving from single core to multicore processors to bring in more parallel processing of workloads. But as the predictability of these workloads became difficult, the usual control flow architecture and algorithms created challenges in executing parallelism. The hardware too was optimized for control flow architecture and single core processors, and the shift to parallelism meant that dynamic resource allocation on the hardware for parallel operations became a challenge.

As these changes were playing out, data flow computing gained popularity along with an explosion of DSAs - domain specific architectures for specific workloads in AI, HPC (High Performance Computing) etc.

‘Morphing’ our ‘Machines’

Morphing Machines is building REDEFINE^TM- a runtime reconfigurable manycore, parallel compute processor which allows the flexibility of an FPGA, but with the specificity of an ASIC – keeping in mind the various workloads and features that companies would want to incorporate into their products in the future. In addition to building the processor, the team is also building a supporting software toolkit and development environment– all of which are integral for the adoption of new processors in the industry.

The team has filed, and has been granted multiple patents over the years, with dozens more in the pipeline to be filed.

Initial demonstrations of the technology on customer workloads have shown extremely encouraging results.

Timing it right

Today we are seeing the fastest rate of technological innovation at any point in human history. The new age of AI demands swift adaptation to the changing market demands, without the need for time-consuming and expensive hardware redevelopment. Software improvements are far outpacing hardware development timelines and capabilities.

BCG wrote a piece in 2021 titled 'Looking to Nature for the Next Industrial Revolution' where they used the term Nature co-design. As the semiconductor industry evolves from hardware-first to software-defined hardware solutions for the new-age requirements, we will see an increased prominence of hardware-software co-design. Hardware will be built to work seamlessly with various rapidly evolving software stacks that are the hallmark of the current demands of AI, High Performance Computing etc.

Morphing Machines’ technology allows it to capitalize on these requirements - a strong desire for new features but also an inclination to not upgrade hardware frequently. Would a customer want the most recent ADAS feature? Absolutely yes. Would they want to upgrade to a new car just because the manufacturer has a new ADAS feature that they cannot update over-the-air? Perhaps not.

The flexibility offered by Morphing Machines’ REDEFINE^TM processor open up a number of market applications predominantly marked by the advent of 6G, ADAS, datacentres, language models, and the need for AI acceleration. The demand for chips in these markets amounts to over a trillion dollars by 2030.

While the industry has been dominated by the likes of NVIDIA, Qualcomm, Intel, ARM and many others, we believe that Morphing Machines’ technology can be complementary to these offerings. In our conversations with various industry stakeholders, we were convinced of the white space of run-time reconfigurability and parallel processing for new-age workloads with ASIC level performance. It is also a project that seemingly the large companies have attempted in the past but the complexity of the technology combined with market not being ready for run-time reconfigurability have deterred many from pursuing this.

Cornami’s raise of over $120M from the likes of Softbank and Applied Materials, Sambanova’s raise of $1B+ from the likes of Softbank, Google, Intel amongst others, and Groq’s hardware-software full-stack support environment approach all serve as directional validation of Morphing Machines’ vision.

Decades of research that is at the cusp of commercialization

Morphing Machines is the brainchild and culmination of almost 2 decades of work led by Prof. Nandy (who recently retired as a Professor at the Indian Institute of Science, Bangalore after 40years there) and Dr Ranjani Narayan (who holds a PhD from IISc in Computer Architecture and later worked at Siemens and HP). The entire work was broken down into small pieces and each has been meticulously and extensively cross-examined by academic and industry experts over the years.

Deepak Shapeti (Stanford University graduate and previously an entrepreneur himself) joined the team as co-founder and CEO in 2021 and has since brought a razor-focussed approach towards taking the tech from the realms of academia towards commercialization. In one of our conversations, Deepak emphasized how not being a semiconductor native has been a huge advantage as it has enabled him to solely focus on building solutions that cater to the needs of the customers and not be charmed by the tech itself.

Bringing a relatively new product offering to a customer has its fair share of challenges - mainly in educating the customer. But as always, the proof is in the pudding. The current fundraise will allow the Morphing Machines team to demonstrate their reconfigurability capabilities through FPGA emulations for various use cases before proceeding for soft IP sales and tape outs. Once the performance is proven, the long-term benefits of reduced design costs and timelines will be a strong value proposition for customers.

Morphing Machines has been a beneficiary of both the DLI (Design Linked Incentive) and the C2S (Chips to Start-Up) schemes by the Govt of India. The incentives are a part of the larger outlay of Rs 76000 Cr (~$10B) reserved for a strong indigenous semiconductor ecosystem.

India has been holding a fifth of the world’s semiconductor design talent with the large fabless semiconductor companies who have their R&D teams here. The time is now ripe for us to build world-class products and lead the charge for the semiconductor revolution across the globe.  

We are super excited to partner with Morphing Machines in their journey to push the frontiers of the semiconductor industry and wish the team the very best of luck.

To join the Morphing Machines team, write to them at info@morphing.in

_________________________________________________________________

We at Speciale Invest believe in supporting breakthrough technologies that have the potential to solve pressing global problems.

As early-stage investors, we like to get our hands dirty early on and support founders in their zero to one journey with patient capital, business development opportunities and hiring. We enjoy and thrive in the risk that comes with backing deep-tech startups at the pre-product stage and help through product-market fit, early customers and scale-up.

If you are building something exciting in semiconductors and the broader deeptech/ deepscience, please do write to us at info@specialeinvest.com

To know more about Speciale’s investments in disruptive technologies, please check our portfolio