When Apple took the wraps off its latest Macs this week, the most interesting thing about the new machines had been well-trailed in advance. They are powered by chips made by Apple itself, a departure after 15 years of relying on Intel.
Where Apple did surprise, though — and even shock — was in the claims it made for its custom-designed silicon. A MacBook Pro running the new M1 chip, for instance, is said to be able to handle 20 hours of video playback on a single charge, twice as long as an equivalent Intel-powered machine.
Those claims have yet to be subjected to independent testing, and moving to a new chip architecture is no small undertaking given the need for a completely new library of software.
If Apple is proved right, it will have gained an unexpected advantage. It will also have shone a spotlight on the new battleground that has opened up among the leading tech companies, as they look to customised chips to handle their most important applications.
Since the PC era, which saw the emergence of specialised chip companies, hardware makers and software concerns, computing has been largely defined by a horizontal structure. Along with the specialisation needed to design ever more complex processors, chipmakers faced the steep capital costs of keeping up with the most advanced manufacturing technology.
A number of things are forcing this deeper level of vertical integration in the tech world. One is that computing needs are changing. The past decade has been defined by the rise of smartphones and centralised cloud data centres.
This is giving way to the age of big data, artificial intelligence and the internet of things. The computing demands are soaring as tech companies look to put AI into every device. Boosting performance while reducing power consumption, always a key issue for chipmakers, has become critical.
In this world, customised chips are the quickest way to a competitive edge. Tesla, for instance, is betting that it can produce a fully self-driving car using nothing more than AI to analyse the images from the car's cameras. Most of its rivals think this is overly ambitious. However, with the equivalent of a supercomputer equipped with chips designed in-house in the trunk of every one of its cars, Tesla thinks it has a shot at proving them wrong.
Nowhere has this been more evident than in the data centres of giant internet companies. Amazon signalled its intention five years ago with the purchase of chip start-up Annapurna and it now makes processors for use in cloud computing centres. Google is already into the fifth generation of its superfast AI chips, while Facebook has been working on something similar.
China’s leading tech companies, fearing the loss of access to US chip technology, have an added incentive to get into the silicon business. Alibaba, for instance, showed off its first AI chip last year, along with one designed to be used in the new equipment required by the internet of things. The latter is based on the open-source Risc-V architecture. Picking technology that is freely available like this — rather than the more widely used designs that must be licensed from ARM — is a sign of how Chinese companies are trying to protect themselves.
Another factor aiding this new verticalisation in tech has been a shift in the structure of the semiconductor industry itself. The emergence of specialised chip manufacturers, led by Taiwan Semiconductor Manufacturing Company, has made it easier for newcomers to produce their own silicon. Apple’s new M1 chips, made by TSMC, rely on a manufacturing process that is more advanced than anything Intel can match.
But there is far more to producing a chip than design and manufacturing. In between are what Matt Murphy, chief executive officer of Marvell Technology, calls the “nitty gritty, nuts and bolts” work of turning a piece of silicon fresh from the manufacturer into a final product, including packaging, assembly and testing. That has led companies like Marvell to seek a new business working alongside the big tech companies.
Google may do much of the work in designing its AI chips, but it relies on Broadcom to supply all this extra chipmaking nous. In the same way, the SQ2 chips Microsoft makes for one model of its Surface tablet are “co-designed” with Qualcomm.
How far companies like these will eventually go in competing head-on with established chipmakers is unclear. But to judge from Apple’s claimed breakthrough this week, a deeper understanding of silicon has become an important differentiator.
Richard.Waters@FT.com
"strategy" - Google News
November 13, 2020 at 12:36AM
https://ift.tt/32zDxAY
Apple chips away at a new strategy for computing - Financial Times
"strategy" - Google News
https://ift.tt/2Ys7QbK
https://ift.tt/2zRd1Yo
Bagikan Berita Ini
0 Response to "Apple chips away at a new strategy for computing - Financial Times"
Post a Comment