Compared to the rest of the auto industry, Tesla Motors (TSLA -1.57%) is a lot more vertically integrated than incumbent automakers. A recent Goldman Sachs report estimated that Tesla is now 80% vertically integrated. In contrast, traditional OEMs focus primarily on engine and powertrain development but outsource most of everything else, essentially becoming incredibly efficient systems integrators.
Is Tesla really pursuing the ultimate step in vertical integration?
Two peas in a pod
Earlier this week, reports surfaced that Tesla had a prominent chip-design specialist away from Apple (AAPL 0.60%). Peter Bannon has now joined Tesla as a director, leaving his position at the Mac maker where he helped design semiconductor architecture. Bannon had joined Apple when Apple acquired PA Semi way back in 2008.
This news comes after Tesla grabbed Jim Keller in January. Keller started PA Semi with Bannon, and together the two led the development of Apple's first A-chips, the A4 and A5. Keller had left Apple in 2012 to join AMD, where he served as the company's chief chip architect. Keller left AMD in September. Keller is now Tesla's vice president of Autopilot Hardware Engineering.
So now Tesla has some truly world-class silicon talent that are both legends in the realm of microprocessors. What could the electric automaker have in store?
All that and a bag of chips
The natural speculation is that Tesla is planning on designing its own chips. When asked directly about the hire on the last earnings conference call and whether or not Tesla could have custom silicon ambitions, Elon Musk expectedly dodged the question with a "no comment." But this is an insane notion.
Designing custom chip architectures is incredibly sophisticated and expensive, and Tesla's unit volumes are fairly modest compared to smartphones and tablets. Tesla sold 50,000 vehicles in 2015. Apple sold 281 million iPhones and iPads in 2015.
Even if you acknowledge that the Tesla Model S has at least two CPUs from NVIDIA (NVDA -1.41%) -- a Tegra 2 in the instrument cluster a Tegra 3 in the touchscreen head unit -- we're still talking about a limited application in terms of volume. Ideally, you'd want to be able to scale those hefty R&D expenses over a large unit base in order to reduce average costs per unit.
The benefits of a custom chip design are also uncertain in the context of an electric car. Apple took control of silicon design in order to tailor performance with power efficiency and take its differentiation to a new level. But microprocessor power efficiency is less critical in the context of an electric vehicle, since it has a massive battery and the primary power consumption comes from the drive unit. And displacing NVIDIA would also generate negligible cost savings, since the Tegra chips that Tesla uses are quite old at this point (it's not clear yet which NVIDIA chips are in Model X), particularly when we're talking about the bill of materials for a car that starts at $70,000.
Tesla is a tech company
It's worth pointing out that Keller and Bannon have extensive skill sets with silicon beyond designing custom microprocessor architecture, so it's entirely possible that Tesla will use those skills in a different, unforeseen way.
While traditional automakers continue to play catch-up in terms of technological features, Tesla could potentially leapfrog itself and broaden that tech gap even farther since it's apparently taking its silicon game to the next level.