If the story is correct, Apple is not only getting "magical" technology, but also some of the world's foremost flash memory technologists. Update: It is official: Apple is buying Anobit for $500 million. End update.
Apple's high gross margins and huge cash hoard mean it can make investments that no competitor can. Many of their investments are in the supply chain - thousands of CNC machines, or billions of dollars of flash - but they also make technology investments where they believe they can create a lasting competitive advantage.
Will buying Anobit for a reported $400-500 million give them a lasting competitive advantage? Here's the scoop.
Flash is not a well-behaved storage medium. It has many problems - fast writes are hard, data wants to leak away, and there's a ticklish relationship between the number of writes and how long data can be retained - and it's getting worse as chip capacities increase.
Flash works by trapping little clouds of electrons in cells. As feature sizes shrink those cells get smaller - and so does the number of electrons in a single bit. With the move to 3 and 4 bits per cell, the electron cloud gets down to a few hundred electrons per bit.
Electrons - tiny, negatively charged particles - like to escape their cells. That threatens your data integrity.
Anobit's technology combines signal processing with flash management to tame rowdy electrons and make flash look like a proper storage medium.
Is that a 1 or a 0? Teasing the answer from ever-smaller electron clouds is hard and getting harder. That is signal processing.
Flash cells interact with each other - for example, voltage levels leak from one cell to another - and cells are sensitive to workload - as the number of writes increases so does the error rate - and these group effects must be managed.
Many flash errors, such as voltage level leaks, behave in predictable ways: voltage always leaks from high to low. Predictable errors can be corrected and then more sophisticated techniques applied to corner cases.
What Anobit delivers
Anobit designs controller chips that make flash behave.
- Reliability. Deep understanding of flash behavior enables Anobit controllers to make flash much more reliable than standard flash specs suggest: they can make 2 level cell flash as reliable as raw single level cell flash is today - a 10x improvement.
- Performance. Flash can be written much faster than normal at the cost of more errors. But if you can fix the errors you can have fast flash.
- Endurance. Reducing errors and managing the flash writes means you can make flash endurance - how long it retains data - much longer.
- Power. Writing flash takes power - 128 flash die can't be simultaneously written on a SATA power budget - and fast writing takes more power. Those trade-offs need management.
The Storage Bits take
If true, the Anobit acquisition is Apple's biggest hardware bet ever. And it is a good bet.
Anobit gives Apple a powerful competitive weapon that can be used to both reduce costs and/or increase performance, while increasing product quality in terms of reliability and battery life. Sure, flash manufacturers have the money to invest in competitive technology, but they'll sell it to all comers, leaving their customers with no differentiation.
And Anobit's expertise - they've applied for over 60 patents - is readily transferable to whatever next-gen technology overtakes flash. This acquisition is a long-term bet on the importance of cheap, fast and reliable solid-state storage for keeping Apple on the leading edge.
Comments welcome, of course. Now if Apple would invest in replacing the much hacked HFS+ file system.
For some details on what Anobit does, check out this StorageMojo post from April. Here's the link to Calcalist that was supposed to be in the overview.