I dont know. My close relative works as infrastructure engineer at major automotive company, and about every 2 years they are buying new computation clusters (think tens to low hundreds thousands of cores, along with network and storage infra). Every time a public tender is held, and every time there are several companies promising to achieve required results with whatever hardware.
Last decade I know of three different firms being selected, where they guaranteed to succeed on speed requirements (some even saying its tested inhouse), only to admit failure after year of trying, taking all their hardware back with no pay. IBM at higher price was selected as backup, and every time was the only firm actually able to deliver on their promises and meet the benchmark requirements.
I often see IBM getting a lot of flak here for whatever reasons, just sharing some counterweight.
It seems EXTREMELY unlikely that Cray would be unable to meet performance requirements and promises. They've consistently done so for decades in all manner of government and private institutions.
No, they've consistently paid penalties for not doing so, especially during the decade or so their network stack tended to implode, taking the shared filesystem with it. They also benefited, during those decades, from an NSA program explicitly designed to fund US HPC OEMs in order to maintain American industry leadership in the space. I think it was a good policy, but it doesn't mean that Cray was hitting home runs right up to the HPE acquisition.
I've sat in on plenty of B2B sales pitches, and am fairly certain that no salesman ever believes that you will actually run those tests. And if even if you did, that you would actually throw away a full year of integration work to go with the backup plan.
Plus, having seen/heard a lot about how salesman are employed - all too often their commission/compensation is locked in when the contract is signed. They get their money regardless of what happens a year or two down the road, and are thus incentivized to "misrepresent" performance and capabilities.
Any sane HPC program has an acceptance testing phase in the delivery terms of the contract, with specific penalties (IBM calls them "considerations") for failing to meet the technical requirements. Sales teams in the HPC are extremely aware of this.
Quantum computers, cutting edge crypto, cutting edge physical reseach just to name a few? I think we are pretty much at the forefront of a lot of things.
disclaimer: I work at IBM Research but this is my own opinion.
BWAHAHAHA! Nothing about that program had any redeeming quality, the fact that everything was based on solidity was enough to confirm this: I knew IBM would be split into 2 companies when the upper brass refused to wise up to the obvious scams like ETH and eventually with Stellar. So much is put behind short term gains selling Z mainframes and empty promises of Watson for everything that nothing gets left over for anything actually innovative.
Tell me, for all the anti BTC propaganda in the training and talking points, it must really have humbled a few people and hopefully got enough people fired to see that the SEC and even Blackrock ended up on our side and not yours.
Instead all IBM has is a place amongst all the other s** coins no one uses with nearly no utility/volume or value to speak of [0]: whereas BTC keeps breaking hashrate ATH.
You guys really dropped the ball on this one, and to be honest the fact that it's a bunch of dinosaurs running that place made it apparent it was never going to live up to it's promise: I just wished the rank and file had taken more effort to make this happen, some of us did try, as so many corps IBM services were willing to pilot-test this tech.
I hate being this crass, I really do, but seriously you guys really screwed this one up.
They will never sell off mainframe. That's the one thing they can reliably tweak to hit their numbers and/or push other software that people don't want as a "bundle".
What is IBM's competitive position? I always assumed they were in the assurance business. I.e. they have well tested systems with close support. Anyone with more insight?
They're one of the few, if not only, technology vendor that is willing to sell something with a written 6 nine's reliability guarantee (e.g. 99.9999% uptime and conformance with advertised specifications/features), or your money back.
No companies in SV, or Microsoft/Amazon, offer a single product that has that kind of guarantee, that I know of.
Maybe they caught on with all the youngsters who call software and AI stuff "tech" instead of technology, and make a clear distinction between the two.