Behind The Scenes of a Competitive Go-to-Market Strategy | What’s Your Edge?

In this episode of What’s Your Edge? we’re joined by Doug Norton, Chief Marketing Officer of Inspire Semiconductor, a company founded in the early days of the pandemic that’s now launching its groundbreaking Thunderbird “supercomputer cluster-on-a-chip.” Doug is a veteran in the semiconductor space. With Doug’s help, we’re going behind the scenes of a Go-to-Market (GTM) strategy and talk about what it really takes to bring a new product to market, and how mapping the right ecosystem, focusing on customer needs, and measuring what matters can make or break a product launch.

Please join me in welcoming Doug to What’s Your Edge?

Doug, you joined InspireSemi at a critical juncture. One of the first things you did was make a strategic decision not to chase everyone, but to focus on market segments that were underserved. We believe that segmentation and customer-centricity are critical to the success of launching a new product.

Can you walk us through your process for mapping the ecosystem, selecting your initial target markets, and how you decided which segments and partners to prioritize?

That’s exactly right, the very first thing was to determine the markets we would – and just as importantly, would not – go after.

Believe it or not, the initial target for what ultimately became our Thunderbird was crypto mining. Our founder saw a lucrative opportunity for some undervalued alt-coins due to the focus on Bitcoin, so he initially created an ASIC for a couple of these algorithms. It was a technical success, but for just a short time, since, unlike Bitcoin, the alt-coins constantly change their algorithms, making an ASIC (Application Specific IC) a dead end. This convinced him what was needed was a very versatile and easy-to-program chip that, of course, also needed to be very energy efficient to profitably mine crypto. Our CEO, who was an advisor at the time, saw the potential for this to be a much broader and more significant device. So they brought me on shortly thereafter, given my background in AI and HPC (High Performance Computing).

My favorite business book is Blue Ocean Strategy. I suspected the right answer was not to chase the bright shiny object of AI like everyone else, since that is a very “red ocean.” Just as important, I have partnered with NVIDIA in the past and knew better than to compete head-on with them.

What I could see happening was the HPC market, who invented and drove the h/w and s/w infrastructure AI now leverages, was being left with few options as all the chip companies rushed to cash in on AI. HPC encompasses important applications used every day in the real world for engineering, science, and manufacturing. Areas like computer-aided engineering, finding and producing energy (both renewable and Oil&Gas), vehicle crash simulation & analysis, genomics, financial modeling, nuclear fission and fusion simulations, climate & weather modeling, and cybersecurity. The underlying need for all of these big math problems is highly accurate math, which requires double-precision floating point (aka FP64). Meanwhile, all the GPU players and AI chips were chasing very low-precision math that GenAI LLMs require (currently 8-bit, moving to 4-bit).

It sounds like you realized you needed to pivot, and once you had clarity on the target market, you were able to lock down the feature set, design, and specs. In your industry, this is often completed years before the product launch. Drawing on your experience, what are the biggest lessons learned so far?

Buy Your Best-Practices Workbook

Never assume anything, especially if creating a product that will take years and millions of dollars to get to market like a high-end semiconductor. Once we knew the market we were targeting, it was important to ensure product-market fit. So, I reached out to my network and did this, with a key decision being which memory technology we should use. Fortunately, most of our choices were exactly what customers and OEMs wanted to see for the HPC market. But including them in the process was important, helping them buy in early. Most importantly, it gave our team, our investors, and our Board of Directors the confidence to move forward with product development.

You also mapped the ecosystem. You recognized that your core customers would be big computing centers, national labs, and OEMs in sectors like aerospace, automotive, and pharma. Knowing what the customer needs and values and how your product delivers is essential to creating the positioning, messaging, and strategy. At what point did you address the value proposition, positioning, and messaging in the Go-to-Market strategy?

Yes, that was the next step, now that the target market and competitors were identified and engineering was off and running. Another nice thing about focusing on HPC was that open-source software is highly valued. I learned early in my career that hardware is just a platform for software. HPC users were the first to adopt the Linux operating system, and 76% of all HPC applications are open source. And it is often these DOE national labs, universities, and supercomputing centers, including UT TACC here in Austin, that work with these applications and can help port them to a new hardware platform like our Thunderbird. And big enterprise companies in markets like aerospace, automotive, energy, and pharma partner with these organizations and draft off their early work.

But we also don’t want to ignore that 24% of software that is commercial, so we have also engaged with some key ISVs in these application areas.

To jumpstart the S curve, we needed to get software development going, so we launched an early access program with select customers. We provided them with simulators and emulators to get going prior to us having actual silicon, and they will be the first to receive the initial Thunderbirds. The ideal scenario is that these orgs will recommend Thunderbird for these apps, or even say they run best on Thunderbird.

Thank you. Having launched numerous products, I know a lot goes into successfully bringing an innovative product to market. It’s a high-stakes endeavor. Research from MIT Professional Programs found that as many as 95% of new products miss the mark. With so much on the line, having a robust Go-to-Market strategy and the right success measures is more critical than ever. Success depends on what happens long before the product makes its way into the market. What we refer to as Upstream Marketing, that is, where customer insights, competitive intelligence, and strategy are addressed to help ensure faster product adoption.

What are the most critical steps you took in the pre-launch phase to support customers rapidly adopting the product, and how did you engage channel partners and technical influencers to build early momentum?

For sure, never underestimate this. From the ecosystem work, it was clear to me that the best Go-to-Market path was to leverage what works rather than try to change it. This meant the initial product should be a standard PCIe server add-in card that the computer OEMs and integrators would sell. Customers don’t typically just buy compute accelerator cards and replace the ones currently in their servers. They buy new servers populated with the new accelerator cards. Plus, Linux has also become the de facto datacenter OS of choice, so the computer OEMs typically ship their servers with Linux.

But the challenge with an indirect model vs. direct sales is that they don’t work for you. You can only influence and encourage them. The key was to engage with the OEMs early on, in parallel with doing business development with our target end-user customers and markets. This let them hear early customer feedback, so they saw firsthand customer reaction to our value propositions. It also built trust, demonstrating we are committed to an indirect channel by bringing them in early. Then there is the financial aspect; the old saying is that salespeople are coin-operated. The server OEMs often lament that a compute accelerator card can cost as much as an entire server, and each server usually has many of them installed. The icing on the cake was that they can get a much higher deal size, with much better mar...