It’s been a little over 9 years since the Silicon Valley venture capitalist Marc Andreesen proclaimed that “software is eating the world.” At the time, his credentials included playing a key role in the invention and commercialization of the web browser before becoming an investor whose portfolio companies included Facebook, Zynga, LinkedIn, Foursquare, Skype, Groupon, and Twitter. In the almost-decade since he staked out his intellectual position, software has created billions of dollars of wealth at the aforementioned Facebook, at Google, and Netflix. As Andreesen’s own list of ostensible world-eaters illustrates, however, it’s unclear (judging from Microsoft acquisitions Skype and LinkedIn, Groupon, or Foursquare) how much his thesis came true. Think back to 1994-2000, and recall the software companies that indeed turned the world upside-down: Akamai, Netscape, Google, PayPal, MySQL, VMWare, and Salesforce.
As those last two companies foretold (as did Andreesen’s second startup, Loudcloud — later called Opsware), the “where” of software was changing from running on a device or customer premise to connecting via the Internet to some vast data center at an undisclosed location. As we look at 2000 to 2020, many of the companies “eating the world” may in fact build software, but their competitive differentiation includes hefty portions of hardware and infrastructure. It’s easy to see that Amazon writes clever software at huge scale, but with dozens of data centers and more than a million employees, many of whom have nothing to do with code, it’s not meaningfully called a software company. Similarly, ByteDance writes clever algorithms to power its TikTok and other similarly addicting services around the world, but absent server farms, there’s nothing happening.
When scanning US-based innovation in the past 9+ years since Andreesen wrote, there’s really not much to see apart from Airbnb and Uber/Lyft: important companies, for sure. Using software to arbitrage everyday people’s capital (at Airbnb, this was true at the outset but less so now) is a game changer, but it remains to be seen if the three companies can be consistently profitable: right now the ride-sharing companies keep losing prodigious sums of money while compressing drivers’ wages and flexibility further each year. As I argued in a journal article a couple years ago, it remains unclear how much Uber’s notoriously bad behavior derived from a crappy culture fostered by the co-founder and how much it resulted from the realization that the entire model could never work. The whole point of taxi medallions was to limit supply to maintain profitable pricing; absent limits on supply, price will race to the bottom of a market.
All of this is a somewhat circuitous way of suggesting that while software is incredibly important, hardware not only matters, but is actually where more interesting things are currently happening. Intel is (to use this year’s most overused word, albeit correctly) an iconic company, on par with AT&T, GM, and Wal-Mart in that it helped define an entire epoch in business history. Former CEO Andy Grove’s paranoia about being disrupted in Christensonian fashion eventually came true: Arm chips, dismissed so easily at their launch given low benchmark performance alongside their power efficiency, were the final blow that broke the boulder of Intel’s industry dominance.
2020 has been a fascinating year on the microprocessor front. Softbank sold Arm to Nvidia to raise cash to help right the sinking Vision Fund ship. In October, meanwhile, AMD bought Xylinx, the dominant player in the broadly-defined system-on-a-chip market. Taiwan Semiconductor Manufacturing Company, meanwhile, is fab-to-the-world, most visibly to Apple, and its stock has more than doubled off its Covid-low in March. Like Apple, Google designs its own silicon, and outsources manufacturing to either TSMC or GlobalFoundaries, AMD’s spun-out semiconductor operation now owned by the Abu Dhabi sovereign wealth fund. Amazon, meanwhile, bought an Israeli chip-design firm in 2015 that now designs the Arm-based custom silicon (reportedly manufactured by TSMC) that powers big chunks of AWS. Facebook appears to rely heavily on Intel, with AI chips rumored to be in development.
Given how few of us work in close proximity to Google TPUs or Amazon Gravitons, Apple’s recent launch of its M1 chip is the first experience everyday people can have of this chip revolution. The M1 is in broad outlines a close relative of the A14 Bionic chip that powers the most recent iPhones. It is available in the Mac Mini and some laptop computers (that, remarkably, are on sale for Cyber Monday). Given that both chipsets are Arm-based, the recent announcement (in July) and release (in November) of Apple products that migrate off Intel processors is extremely significant. The performance gains are simply staggering: my favorite is that the M1 can emulate an Intel chip — relatively efficiently, but not as fast as Apple’s Arm-native code — faster than an Intel can run at full throttle. This performance comes with battery life measured in days rather than hours, without cooling fans, and at low price points. The fact that TSMC can produce 5-nanometer traces while Intel is announcing delays in its 7-nm products, out into 2022, helps explain the disparity.
It’s unclear what markets Intel will have left: Apple is phasing out Intel chips, desktop computing is in the midst of a steep decline, and Intel captures little of the tablet market. The Wintel duopoly, meanwhile, is no longer an oligopoly, and Microsoft is focusing its energies on the cloud rather than the desktop. Lots of laptops were sold to support Covid-driven work-from-home initiatives, but that blip in sales will most likely not last given general economic softness and the size of this sudden (and one-time) refresh cycle.
I’ll spare my readers the fanboy-like praises of the many reviewers (here’s one summary) but it did warm my heart to remember seeing technology breakthroughs on a regular basis: the Mosaic browser, Altavista then Google search, Keyhole EarthViewer, Gapminder Trendalyzer, YouTube, the iPhone, then . . . what? The iPad, EarPods, Apple Watch, endless meal-delivery apps that will bankrupt already-drowning restaurants — none of these really count as anything that engages me the way the M1 reviewers report feeling. Just to take one example, opening the lid of an M1 Mac laptop has become a kind of sport, given the instant screen readiness. Even browser windows and simple operations are said to snap in a way that redefines the computing experience, reminding a certain kind of user why we went into this business in the first place.