How do we think about building products in Deep Tech?
This week we think about models for understand Deep Tech product phases as they relate to the organisation, the market, and the people involved in each. Plus a round-up of signals to watch.
What’s the big idea?
Hello from the cold and grey extremes of Seattle. Over the last few weeks I’ve been meeting with peers from the Quantum Computing and AI sectors. A common challenge we’ve been discussing is how to think about the progress of products and services in the Deep Tech realms. This week’s feature digs into this topic, exploring some of the key frameworks that are our industry standards. I’m also sharing my own Frontier Tech Phase Model that thinks about the journey “from Science to Technology to Engineering to Product”. Check it out and let me know what you think, as this is an evolving concept.
🔗 How do we think about products in Deep Tech? A suggested framework.
Quantum Book Club: Helgoland by Carlo Rovelli
I read every quantum book I could get my hands on when I joined Quantum Brilliance. But you don’t have to. Here’s the very first of a curated series that kicks off with Italian theoretical physicist Carlo Rovelli’s “Helgoland”. It starts the journey on a stormy night on a remote North Sea island where a young Werner Heisenberg makes a crucial breakthrough in the creation of quantum mechanics. If you’re a fan of Rovelli’s writing like I am, I also recommend catching him on Krista Tippet’s On Being podcast, where his exploration of the moment to moment nature of reality is poetic without going full Deepak, and optimistic without reaching peak Kaku.
🔗 Quantum Book Club: Helgoland by Carlo Rovelli
Signal from Noise
In terms of the signals that have proved interesting this week, we’ve got a look across the emerging open source battlefield for AI platforms, the steady growth of community-centric funding models that seem to be diverting talent away from the legacy startup accelerators, and a bunch of earnings call and investment analysis to give us a feel for how 2024 is going for Deep Tech cash monies. It’s also the first newsletter sent since migrating to Substack, so hello to any new readers joining us!
“Meta man shoots at LLM hype”
In case you missed it, the video and slides are now available from Meta’s Chief AI Scientist Yann LeCun’s 2004 Lytle Lecture at the University of Washington. The talk is titled “Objective-Driven AI: Towards Machines that can Learn, Reason, and Plan”, and the TLDR is LeCun’s perspective that LLMs are overblown, AGI is nonsense, and open source is essential to AI strategy. We also get a view into his progress on Joint Embedding Predictive Architecture (JEPA) (eg this to this) and context around Meta’s AI strategy. Bonus points if you can spot me in the crowd.
“Le chat est mort et pas mort”
Speaking of the French Deep Tech ecosystem, the final agenda for Q2B Paris has been released. Q2B is the leading quantum computing conference for the business side of the sector. While I typically go to the Q2B Silicon Valley edition, the French side is interesting to discover the European quantum companies with earnest aims for the technology. There’s also the usual government representation, and the McKinsey and BCG talks that tend to seed the market sizing and industry discovery and investment reports that we go on to quote over the rest of the year. Yes even me.
“Mistral goes more open than open”
Mistral AI has launched a new flagship model to take on Google and OpenAI. And a surprise partnership with Microsoft that looks to get EU scrutiny. You might remember their seed round announcement, which raised eyebrows. Then the follow-up $400M that valued them over $2B. Their pitch deck hit the mark anyway and the team quickly shipped the Mistral 7B model in September of 2023, published their paper in October, and open-sourced it under an Apache 2.0 license. It’s worth looking deeper when a technical French team tackles an open source project (such as this one and this one).
“IonQ aims to be Nvidia, Cisco, OpenAI in one”
Still on the business side of Quantum Computing, today saw the IonQ earnings call. I’ll be doing a deep-dive on this and other quantum stock in a few days, but the takeaway here is that the wild speculation and retraction of this segment of the market seems to have abated. Even Rigetti’s close call with delisting has evened out after joining in the layoff trend. IonQ is notable for getting ahead of both the quantum funding winter and the end of the SPAC era with a staggering $2B listing. Here’s the presentation from today’s call.
”Ride the (D-)Wave of quantum learning”
On the education side of things D-Wave has announced a starter course for quantum computing training. This comes in the form of two paid courses with an additional upsell to roll the training into D-Wave’s Leap quantum cloud service. Training programs are table stakes for quantum players where it’s less about “bridging the skills gap” as claimed, and more about mindshare and market capture. IBM’s Qiskit community is unmatched in this area, although Q-CTRL’s Black Opal is notable, especially as they focus on the student market. Vendor-specific system training could be a loss-leader, especially with such large scale FAANG layoffs unlocking talent, but quantum startups are not known for their velocity or market reactivity.
“Deep (Tech) in the pocket for 2024”
Rounding out investment chatter, after my recent work on due diligence with a number of venture funds I can attest to Crunchbase’s jaunty “things are down but not as bad as we might think”. What isn’t talked openly about (or announced to tech media) is the role that national strategy is playing to support technologies of interest to sovereign capability. The APAC and ASEAN regions are one example, where Singapore’s role is noticed if misunderstood as a significant portion of NVIDIA’s revenue in 2023. This trend isn’t limited to Deep Tech, but right across the supply chain, and is worth understanding where gov dollars are substituting for venture dollars, and how that changes product development and delivery dynamics.
“More like a decelerator, amirite”
Also on the capital side of things, it’s been a curious few weeks for accelerators. On the same day as reports of Techstars shuttering their Seattle accelerator program, a former Managing Director posted an earnest account of “what went wrong at Techstars”. It reads as a universal summary of issues experienced by many accelerator programs, including Australia’s routinely contentious BlueChilli, or France’s NUMA, which confused the ecosystem even before troubled global expansions fell over. Other Techstars alumni shared their views and experiences: chasing corporations rather than startups doesn’t work. The Techstars CEO thinks otherwise and clapped back via LinkedIn in a manner that gives clues to outsiders what those brutal Glassdoor reviews were on about.
“A modern day Xerox Parc”
At the other end of the scale we see the accelerator model evolving away from startups and more towards talent and community signals. This week Aditya Agarwal reminded us that South Park Commons is accepting applications for their Founder Fellowship (which features up to $1M in funding). The community aims to be “a modern day Xerox Parc”, a favourite reference for Silicon Valley innovation nostalgia. A similar investment thesis has been embraced by the likes of Antler and OnDeck (the latter with varying levels of success and sentiment), and others sure to follow given the displaced talent after FAANG layoffs.
Question of the week
What resources or frameworks do you use when thinking about or evaluating Deep Tech? Hit reply and I’ll share your suggestions in the next newsletter.
I welcome another Product and Tech Newsletter.