NVIDIA is a long-term bet against Moores’ Law
The other day, whilst chatting to a friend (over Zoom this time, rather than coffee), I mentioned that I see NVIDIA as a long-term bet against Moore’s Law,1 which my friend found quite interesting.
After the call and a bit of pottering I realised that NVIDIA is also a great example of how strategy in an ambiguous environment (as we have today) doesn’t rely on strategic foresight. Moreover, foresight is often wrong, and sends us down blind alleys and into dead-ends.
This is somewhat at odds of how most analysts ascribe successes such as NVIDIA’s to a combination of strategic foresight, technological innovation, and adaptability. I would even argue that technological innovation (which is usually read as research-driven development) isn’t as important as we like to think, as innovation tends to bubble bottom-up (from learning by doing) than top-down (directed research).2
To be clear, by Moores’ Law here I mean the mid-nineties when process improvements were doubling computer chip performance roughly every 18 months (about 60% annual improvement). Today, performance is growing at 10-15% per year (doubling in 7-10 years).
The technology industry could see this slow-down coming, and experimented with various forms of parallel processing to keep the relevant numbers going up. Today Moores’ Law is a design goal more than anything else,3 and reflects how many chips we can cram in a multi-chip processor, how many processors we can squish onto a board, and how many boards we can stack in a data centre, than anything else.
While industry has largely focused on propping up CPU performance, NVIDIA acknowledged the impending slowdown. Moreover, NVIDIA realised that a long-term bet on the end of Moores’ Law would pay off. They saw that the future would belong to parallel processing, regardless of the tasks being performed.4 This wasn’t a question of predicting particular future applications, but about recognising a fundamental shift in computing architecture and heading in the same direction.
Strategy is largely a question of direction. As Roger Martin is famous for pointing out, a strategy is not a plan.5 It’s not just a series of actions; it’s a deliberate choice on the direction to head. Without a clear sense of direction, actions become reactive and disjointed.
NVIDIA’s strategy was not a plan to dominate gaming, then crypto, then AI.6 It was the intention to build the most powerful parallel-processing platform possible. The slow-down in processor improvement would then drive use cases to their platform. Hence, NVIDIA is a long-term bet against Moores’ Law.
NVIDIA didn't have a detailed plan for how each market would emerge. Nor had they developed foresight into which markets would emerge, and when. The firm’s technical innovation was also more bottom-up than top-down. CUDA, for example, while a signifiant investment, was an investment in making SIMD more developer-friendly to foster bottom-up innovation, rather than an investment in a predicted top-down market opportunity. One might say that NVIDIA invested in optionality. The firm wasn’t locked into a single application. If one market faltered, the hardware could be repurposed for another.
With CUDA and their software libraries they made it relatively easy for developers to adapt GPUs to new tasks. By investing in developer tools, documentation, and community building, NVIDIA fostered an ecosystem7 that could explore and exploit new applications. They also created partnerships with many key players in the technology industry, which has helped to expand their market reach, and create a strong demand for their products.
This enabled NVIDIA to rapidly respond to market shifts. When crypto mining surged, NVIDIA quickly ramped up production and marketing to capitalise on the demand. Similarly, they were early to recognise the potential of AI and invested heavily in AI-specific hardware and software.
In essence, NVIDIA's strategy can be seen as:
Build a flexible and powerful platform (GPUs and CUDA).
Foster an ecosystem that can explore and exploit new applications.
Execute rapidly and efficiently on emerging opportunities.
This approach aligns strongly with the central ideas in Mastering the Puzzles of Transformation. NVIDIA’s strategy is built on an underlying structural shift, the observation the slow-down in processor improvement would drive use cases to their platform. By building a flexible and powerful platform (GPUs and CUDA), NVIDIA fostered an ecosystem that can explore and exploit new applications. The firm then executes rapidly and efficiently on emerging opportunities. This is akin to recognising structural shifts, navigating complexity, and fostering collaboration. NVIDIA’s success demonstrates the importance of systems thinking and ecosystem engagement to drive systemic change and create value.
Moore, G.E. “Cramming More Components Onto Integrated Circuits.” Proceedings of the IEEE 86, no. 1 (January 1998): 82–85. https://doi.org/10.1109/JPROC.1998.658762 and Moore, Gordon E. “Progress in Digital Integrated Electronics [Technical Literature , Copyright 1975 IEEE. Reprinted, with Permission. Technical Digest. International Electron Devices Meeting, IEEE, 1975, Pp. 11-13.].” IEEE Solid-State Circuits Society Newsletter 11, no. 3 (September 2006): 36–37. https://doi.org/10.1109/N-SSC.2006.4804410.
The classic reference for bottom-up vs top down innovation is Hippel’s Democratising Innovation. For an industry wide focus, Sovacool’s The Importance of Open and Closed Styles of Energy Research is good. Hippel, Eric von. Democratizing Innovation. The MIT Press, 2005. https://doi.org/10.7551/mitpress/2333.001.0001. Sovacool, Benjamin K. “The Importance of Open and Closed Styles of Energy Research.” Social Studies of Science 40, no. 6 (December 2010): 903–30. https://doi.org/10.1177/0306312710373842.
Waldrop, M. Mitchell. “The Chips Are down for Moore’s Law.” Nature News 530, no. 7589 (February 11, 2016): 144. https://doi.org/10.1038/530144a.
Jensen Huang's vision regarding parallel computing, particularly through NVIDIA's GPUs, has been a consistent and evolving theme. Some key aspects include:
Early Recognition of GPU Potential: From NVIDIA's early days, Huang and his co-founders recognised the potential of GPUs to go beyond graphics and revolutionise general-purpose computing. This involved understanding the power of parallel processing. The development of the RIVA 128, as highlighted in the Acquired podcast, showcases early focus on accelerated computing.
CUDA and the Shift to General-Purpose Computing: The development of CUDA (Compute Unified Device Architecture) was a pivotal moment. It allowed developers to harness the parallel processing power of NVIDIA GPUs for a wide range of applications beyond graphics. This fundamentally changed how computing tasks could be approached, especially in areas like scientific computing and AI.
Parallel Processing and AI: Huang's vision has been instrumental in driving the adoption of GPUs for deep learning and AI. The massive parallel processing capabilities of GPUs are essential for training complex neural networks. His statements emphasise that the future of AI is deeply intertwined with advancements in parallel computing.
Energy Efficiency: Huang has also been vocal about the energy efficiency of accelerated computing, stating that GPU acceleration uses less energy than general-purpose CPUs to achieve the same or higher levels of performance.
Vision for the future: His current vision is to continue to push the boundaries of what is possible with parallel processing, and to make AI and accelerated computing available to everyone.
In essence, Huang's vision has consistently centred on unlocking the potential of parallel processing to solve complex problems and drive innovation across various industries.
Harvard Business Review. “The Difference Between a Plan and a Strategy.” May 3, 2023. https://hbr.org/podcast/2023/05/the-difference-between-a-plan-and-a-strategy.
Wu, Andy. “Nvidia.” Harvard Business School Case Case 725-383, October 2024. https://www.hbs.edu/faculty/Pages/item.aspx?num=66542.
For more on modern ecosystems, see my collaboration with Giselle Hodgson. Strategy and the Art of the Possible. Evans-Greenwood, Peter, and Giselle Hodgson. “Strategy and the Art of the Possible.” Deloitte Insights, July 6, 2022. https://www2.deloitte.com/us/en/insights/topics/strategy/business-ecosystem-strategy.html.