As so often happens, three disconnected observations turned out to have a common theme. In order, I’ll connect the Twitch game-streaming platform, local real estate, and the reported decline of corporate R&D to the firm that has remade multiple landscapes. It turns out this is a timely discussion. As Congress and the executive branch debate whether and how to regulate the tech sector, they need to think deeply. What we are seeing is fundamentally new, and thus requires careful attention to the details of governmental oversight. Merely saying “Jeff Bezos is too rich,” or “Amazon is too big and therefore bad” may be politically popular, but it’s intellectually and economically sloppy. We need to develop new tools of analyzing social good and social costs, and new language to weigh the ratio of one to the other. In this age of hot takes and polarized discourse, that’s probably an impossible ask. Call me an idealist and I won’t disagree.
Twitch
Alongside Facebook’s acquisition of Instagram, Amazon’s purchase of the Twitch gaming platform was one of the smartest pieces of M&A I’ve ever seen. (Pop quiz: what is the _worst_ acquisition ever? My nominees will appear below.) The MIT professor T. L. Taylor makes an extremely important point about game streaming: TV used to be something people watched. Game streaming -- and I’m arguing, online video more broadly considered -- has transformed that relationship. Twitch is something people can watch, in a variety of ways, but more important, it’s something they _do_. Game commentary is arguably the most profitable category on YouTube, and Twitch has become a powerful niche site where gamers can watch, play, comment, interact with both gamers and watchers, or opine on non-game issues. Game designers are building to the platform, integrating the gamer community in both development and competitive aspects of their products. Communicating with others as they watch gameplay dates back to the hangers-on around arcade games, and the scale of the audiences in this virtual arcade are staggering: according to Twitchstats.net, at peak load about 140,000 channels (streams) are reaching 3.8 million viewers.
As a student of mine pointed out, meanwhile, Amazon won broadcast rights to Thursday NFL games in the league’s most recent media auction. He speculated that Twitch will be a facet of the Amazon streaming of the games. This could involve fantasy sports, snarky in-group commentary, and/or actual wagering at some point. It’s getting tough to track all of Amazon’s forays: pharmacy, grocery, drones and logistics, the MGM acquisition, the fast-growing ad business, cloud computing. Michael Porter's 1980s business school concept of an industry really doesn’t apply any more.
Urban archeology
I was strolling through downtown Syracuse earlier this week. A massive space used to house the flagship Dey Brothers department store from 1894 until 1992. Like many other downtown buildings, it hasn’t found its 21st century purpose yet: an art center seems to be at least one tenant, but the bricked-over windows and giant rusting awning over the main entrance highlight the contrast between the structure’s past vibrancy and today’s urban uncertainty.
In the 1950s -- sources disagree on the exact year -- Dey Brothers joined its main competitor Addis as early tenants in the Shoppingtown mall in the adjoining suburban town of DeWitt. Shoppingtown was originally an open-air mall but not surprisingly (given the Syracuse climate) converted to an enclosed space that celebrated a grand opening in 1975. In 1974, meanwhile, a 75-store mall opened about 3 miles further east of Syracuse in Fayetteville. It featured a carousel built in Europe, and Addis and Dey (they’d merged) opened a $7 million store in the 1990 expansion. That didn’t last long. Yet another mall, then named Carousel, opened to the west, closer to Syracuse proper.
By 1993 all Dey Brothers-related stores had closed. Fayetteville Town Centre was partially demolished and converted into an open-air mall. Shoppingtown currently sits empty, though a multi-use redevelopment plan was just announced this week. Carousel, by now the largest mall in the state of New York, was renamed Destiny USA by its developer Pyramid Companies, a huge mall operator in the Northeast US. Its $285 million in bonds were downgraded to junk status last month.
What does Syracuse real estate tell us about the future of retail? Amazon just opened two facilities here: one is a 3.8 million square foot distribution center in Clay (a northern suburb). If that number sounds big, it is: the only larger industrial structures, apart from similar Amazon buildings, are Tesla’s Nevada battery plant at 13.6 million, and the Boeing facility in Everett, WA where they used to build 747s. The other local Amazon facility is a 112,000 sq ft “delivery station” in DeWitt, the same town that boasts a currently empty Shoppingtown mall. That building will host a fleet of local delivery vehicles, standing as one of more than 500 such Amazon sites, according to one consulting firm’s estimate. It’s a neat inversion: Shoppingtown was built for cars to converge on the mall, whereas a delivery station is designed for [mostly electric] delivery vehicles to disperse into the community.
The decline of corporate R&D?
A colleague forwarded me a paper out of Duke’s Fuqua school of business. Entitled “The changing structure of American innovation: Some cautionary remarks for economic growth," the paper analyzes the decline of corporate R&D as measured by PhDs hired, patents, publications, and other measures. There's a decided lack of appetite among corporate shareholders and managers for maintaining modern-day successors of GE Research, Bell Labs, Xerox PARC, or IBM Research. (A high-profile example: under pressure from an activist investor, DuPont shut its R&D lab in 2016, despite a distinguished history of publication and even a Nobel Prize.) Universities were theorized as taking up the slack, but the paper casts doubt on the efficacy of this approach.
Although Google gets considerable attention in the article and Microsoft's commitment to publication is highlighted, Amazon is barely mentioned. Given global competition, neither publications nor patents are obvious activities for a company making serious gains in data center management, algorithmic efficiency, machine vision, codebase maintenance, and many other areas. That is, I think private-sector R&D is delivering concrete innovations but they aren't being announced or shared in traditional channels. Google's Tensor Processing Unit chips, optimized for machine-learning tasks, come to mind. Amazon has developed its Graviton, designed for data-center efficiency, in the same custom silicon space. Both chips are proprietary.
Cloud computing hardware has evolved to the point where racks of specialized devices, not general-purpose servers, are the unit of analysis. While much of this material is open-sourced, it is not included in the article's critique of corporate R&D. (See ocp.org) Finally, huge advances in algorithmic science have been achieved, often by academics hired by Facebook, Google, Amazon, and elsewhere, but they aren't always made public.
This migration of academic talent into industry is fascinating for many reasons. Money is obviously a factor: Uber hired away essentially an entire Carnegie Mellon autonomous vehicle lab -- 40 people -- in 2015. But resources are also a draw: instead of spending weeks or months writing grants to get data sets, computing time, or post-docs, researchers inside the big platform companies have access to effectively infinite computing power, no teaching expectations, no committee work, and no career-dependent need to publish in journals with acceptance rates in the low single digits. The Duke paper quotes an astonishing yet representative statistic: inside Google, researchers use the JFT-300M dataset (300 million images, 375 million labels). Academic researchers aspire to use data sets similar to Stanford's Imagenet, which had at last report about a million images. Which setting will generate big-data insights more readily?
While I think the Duke authors are onto an important question, the differences between software and hardware or chemicals are not sufficiently addressed, I don't think. Facebook can roll out an algorithmic refinement tonight; Bell Labs or IBM faced a 3-5-year lag, at least, between lab insight and commercialization. Similarly, when China's "Big 3" of Baidu, Alibaba, and Tencent (soon to be joined by ByteDance) operate under a different set of financial, subsidy, regulatory, and ideological assumptions, global competition plays out differently than GM vs Volkswagen or Dow vs BASF did back in the day. National patents don't make the competitive sense thy once did.
*******
Given the many facets of just one company, talking about "Big Tech" as some kind of homogeneous oligopoly doesn't make sense. Just addressing advertising (and therefore privacy) alone, Amazon, Apple (yes Apple), Facebook, and Google are all intense competitors, operating in a "Wild West" environment due for some rules of the road. Comcast, Verizon, and another dozen companies deserve scrutiny in that same discussion. Moving to logistics, Amazon competes head to head with FedEx, UPS, USPS, and others. Have that discussion too, on its own particular merits, possibly asking pointed questions about universal service, subsidies, etc. Then do cloud computing, maybe addressing data portability, with Amazon, Google, IBM, and Microsoft in the room. If private-label retailing is a concern, of course Amazon is involved, but so is every major grocery chain, many apparel companies, and Target and Wal-Mart. The point is simple: address corporate competitive issues in specific terms rather than with cheap and empty posturing.
(My personal, completely unscientific hall of shame of bad acquisitions: GE/Alstom, Microsoft/Nokia, eBay/Skype.)
Posts
