A motorcycle parts company in Southern California in the 1990s used the phrase as a sales slogan for oversized Harley Davidson exhaust pipes. Others say the phase originated with mogul skiing in the 1980s as skiers taunted each other into more and more spectacular runs. Or maybe it came from the surfing and refers to the riding really big waves…
The group of manufacturers that can afford to build microprocessors fabrication plants is small. Modern fabs and production tools cost billions of dollars to build and the upfront investment is very high. It’s down to Globalfoundries, Taiwan Semiconductor Manufacturing Company (TSMC), Samsung, and Intel. Intel mostly makes its own chips, but has begun to do foundry work for others. Samsung is a hybrid, making some parts for itself and some for others.
TSMC is a pure foundry and dependent on a tiny number of customers to drive the bulk of their revenue, with ~16% of TSMC’s revenue coming from Apple and 16% from Qualcomm. In the old days, graphics cards from AMD and Nvidia helped drive the early adopter cycle, but that hasn’t been true for several years. Mobile devices, not GPUs, are driving foundries to newer process nodes. TSMC announced it will build a $15.7 billion foundry facility focused on the upcoming 5nm and 3nm process nodes.
GlobalFoundries’ Fab 8 (see picture below), one of the largest semiconductor fabs in the United States, sits on 233 acres of the Luther Forest tech campus in Malta, N.Y. carries the DNA of chip giants such as AMD, IBM and Samsung. Everything about the place is gigantic, from the fields of supply tanks full of different gases out back to the network of pipes that carry them into the plant. The clean space itself is the size of six football fields. Investment in Fab 8 — underwritten primarily by Mubadala Development Corporation, an arm of the Emirate of Abu Dhabi — is $10 billion and climbing.
As foundry node shrinks have become more difficult, fewer and fewer companies have attempted them. As of 2011 (see diagram below), there were 19 companies with production capacity at 130nm and just four companies at 20/22nm (GlobalFoundries eventually joined this group, bringing the total to five). There are currently four only foundries offering 14/16nm technology — Samsung, TSMC, GlobalFoundries, and Intel.
There has been much press recently about Oracle launching its own public cloud. But like the giant chip factories, public cloud requires massive investments to achieve economies of scale. Amazon, Microsoft, and Google are positioned to make those investments but most other players are not. Companies closely guard any direct insight into their data center infrastructure (Google transacted their investments under The Salinger Group for instance) because you can infer how big they are from knowing data center capacity.
Rachel Stephens from RedMonk published an analysis last summer that uses information about size of companies (based on publicly available financial records) to make inferences about their ability to compete in the cloud market. Building out data centers is capital intensive, and these sizable investments are generally encapsulated within the balance sheet as a long-term asset under property, plant and equipment (PP&E). Note that nearly all of these companies have other major lines of business that obfuscate the impact of their cloud strategy in their consolidated financial statements.
The waffle charts below show how net PP&E balances have changed for each company between 2014-2015. IBM is notably absent from these graphs, as they experienced PP&E decreases in each year. CenturyLink has attempted to pivot to a cloud business model but their significant PP&E balances reflect their core businesses in fiber/telephony/TV. HPE announced in October 2015, that it was shutting down its public cloud offering. Rackspace is now focused on managing the cloud platforms of others vendors instead of its own cloud offering.
Competitiveness in the cloud market is largely reliant upon infrastructure investments. From 2014 to 2016, Amazon, Microsoft, and Google have invested billions into expanded infrastructure. The value of Amazon’s net PP&E has increased by 204% ($15.6B) since 2013; Google’s increased by 145% ($17.9B) and Microsoft’s by 83% ($7.6B). All three companies have both internal business drivers and incoming revenue streams to subsidize further data center expansion. The drivers for Google are search and applications such Gmail, etc. For Microsoft, it is Office365, and for Amazon it is www.amazon.com and all its associated services such as Amazon Prime. Those investments occur even before considering the profitability of their public cloud offerings. The trajectory, size or both of the other competitors’ investment raises questions about the number of serious participants the public cloud market can reasonably expect moving forward.
Gartner recently published (see below) its annual Magic Quadrant for Cloud Infrastructure as a Service, Worldwide. The report shows same top three players in the public cloud market but ranking does not reflect the relative size of their infrastructure investments. Amazon continues to dominated the market having an installed base many times greater than its other competitors combined. Microsoft has been able leverage its enterprise relationships and more modest infrastructure investments to reach second place. Google Cloud does not have the same breath of features and services of the other two marketplace leaders but its global infrastructure is the most robust and capable.
Gartner also notes that rescaling of the Ability to Execute axis to reflect the expansion of customer requirements and unmet needs in the marketplace. AWS appears to be less far up the Ability to Execute axis than it did in the previous year, despite significant improvements in its business in the past year. Other public cloud vendors may substantially improve their capabilities from year to year, yet not achieve significant movement in their position, because their position is relative to the overall market.