Intel loves the 'internet of things' – if the 'things' have Quarks and Atoms in them, anyway

What will Chipzilla eat?

Rule #1 of being a bit corporate is to understand how your operation fits in the market. One of the keys to this is "predator thinking".

Predator thinking means you need to need to fit your corporate metabolism to the market prey you seek.

A lion chases zebras, not ants or elephants. A hover fly chases aphids, not zebras.

A small consulting company might chase $100k contracts. A large corp doesn't get out of bed for opportunities worth less than $20M.

For its more recent existence, Intel has been completely x86 focused and pretty much only sells high margin chips. These are very expensive to develop, but then the prices are high (relative to other chips) and the payback is huge. Intel's corporate metabolism is structured to play this high-margin game well.

Embedded systems typically depend on very low-cost chips and this is especially true of the bottom end devices which will dominate the IoT (if it ever really takes off).

When we're talking about internet connected lightbulbs we're talking about devices that will have less than 50c of electronics in them. That is something you can do with 8-bitters and, perhaps at a stretch bottom-end ARM devices. It is not something you can do with an Atom.

Remember, it isn't just the CPU cost that matters. It is also the associated power supply etc. A low-power Atom needs a very low ripple power supply which is large and expensive. An AVR (or similar) can work with a voltage from 1.8-6V (or even wider) and any level of ripple you throw at it. That just needs three or four passive components that cost less than 5c all up.

But back to the corporate metabolism... Intel is geared around high margin chips making many dollars per chip. Can they re-jig the company to be able to be profitable on low-margin devices that make them a penny or two?

"Intel's corporate metabolism is structured to play this high-margin game well."

Yeah, that's why IA64 took the 64bit x86-followon market by storm?

"Can they re-jig the company to be able to be profitable on low-margin devices that make them a penny or two?"

How would that work? ARM succeeds by working closely with an increasingly wide network of chip design and fabrication partners who cover a huge number of market sectors, from safety critical self checking microcontrollers for underbonnet use, to pocket rockets for phones and (maybe) datacentres. Plus a good chance of one or more in every piece of consumer electronics on the market (maybe half a dozen in a desktop PC). In a PC, the ARM revenue is a lot less than Intel's. In every other bit of electronics, the x86 revenue (and hence Intel's total revenue) is negligible.

Intel has historically succeeded by working closely with a network of PC assemblers (x86 desktops, laptops, servers) who themselves mostly work closely with Microsoft. Intel have repeatedly tried to get out of that niche but I can't think of any notable successes and there are plenty notable failures. That system assembler network is shrinking (no CPQ, no IBM, Dell on the way out, Microsoft losing the plot too, and the sales figures show the x86 market is in a very poor state). ARM everywhere has changed the rules of the game.

It would take Intel years to catch up. If they could.

Look back a few years.

Intel had a chance to back ARM (in particular, StrongARM) and be a leader in the non-x86 low end market. Their management ignored it.

Intel. The x86 company. With an increasingly irrelevant future.

Sell.

SCADA called, the cease and desist is in the mail.

all the HVAC (heating, ventilation, and air conditioning) units - blah blah - gateway that would communicate -blah blah - central gateway, - blah blah - coordinating gateway - blah blah - interoperability of the entire enterprise.

Gee, what a swell idea, controlling stuff with computers. Why did no one think of this before ...

Intel seen to have forgotten to talk to IoT providers.

We've been building these systems for many years.

Itty-bitty ARMs at <50 cents each are used for the myriad of end devices from disparate manufacturers, the small number of edge "servers" run on MIPs or mid-range ARMs and finally, the core, campus-wide, machines are Atoms.

In other words, the Internet of Things requirement for Atom-class devices is roughly two per campus (main & backup).

The place we use Intel the most is actually GUI interfaces - large touchscreens, and the building-status widgets on the normal office PCs.

Intel needs to bring the cost down a lot to stop the large touchscreens drifting towards ARM and MIPs - as the panel prices fall and ARM GPUs get better, it gets harder to justify the price and power budget of an Intel SBC.