If memory serves, AMD hasn’t bought much of anything since it impoverished itself buying Canadian graphics chipmaker ATI Technologies for $5.4 billion back in 2006. It’s certainly the first major acquisition since AMD CEO Rory Read joined the company last summer.

The acquisition is supposed to be a strategic investment in the type of dense, low-power, high-bandwidth platform attractive to the cloud set and spook agencies these days. And that’s an itch ARM also means to scratch.

The sweet irony of the situation is that the only gear SeaMicro is currently selling is based exclusively on Intel parts, to wit the low-power Atom chip, and just recently a low-power Sandy Bridge Xeon processor, which lets it run heavyweight widgetry like Hadoop, real-time analytics, Java apps, PHP, Memcached and NoSQL in addition to webby stuff.

However, the industry’s newest couple says it’ll have an AMD Opteron-based model of SeaMicro’s widgetry out in the second half.

They didn’t identify the chip they mean to use, but it could be the upcoming one-socket processor code-named Delhi with four to eight cores that was the only future chip that fits the SeaMicro timeframe to be clearly targeted at microserver use on a recent roadmap Read gave Wall Street.

It would then appear that Intel, which calculates that microservers will represent 10% of the overall server market by 2016, has lost its sole microserver champion in the move.

CEO Andrew Feldman says that SeaMicro, which will be run as a free-standing unit, will continue to support its Intel line for the life of the chips it’s using, which he said was another two years. Intel in turn said two years is a long time and that another Intel flag carrier was bound to turn up.

Be that as it may Feldman said the conversation with Intel was “hard” without giving up any of the doubtlessly juicy details.

Among other things Intel hasn’t been able to break ARM’s vise-like grip on the lucrative low-power mobile market and by the end of the year the ARM contingent should have samples of four-core 64-bit silicon that’ll give it a leg up in the server and desktop market. Intel, meanwhile, will have a next-generation Atom.

AMD has also appeared to be leaning toward adopting the AMD chip, which SeaMicro rival Calxeda has developed into a quad-core 32-bit Server-on-a-Chip (SoC) that HP means to sell in its own so-called Redstone microservers for which it has set elaborate plans in train.

John Fruehe, director of product marketing for server products at AMD, acknowledged that AMD might eventually graft an x86 chip and an AMD chip together onto the same SeaMicro SoC. It also has its ATI-derived GPUs to play with, he said. There could also be plain vanilla ARM boxes.

See, SeaMicro’s so-called supercompute fabric, which connects thousands of processor cores, memory, storage and input/output traffic, supports multiple processor instruction sets.

AMD said the acquisition will accelerate its disruptive server strategy so it can stake out a “data center leadership position.”

One of the move’s disruptive features is the fact that for the first time AMD will be competing with its own customers while at the same time hoping to turn them into SeaMicro OEMs pushing the widgetry into the cloud data centers that IDC projects will be the fastest-growing segment of the server market through 2015.

Supposedly HP was supportive when notified of the acquisition. Guess we’ll see.

Feldman, who will become general manager of a new AMD Data Center Server Solutions business reporting to Lisa Su, AMD’s new general manager of global business units, said SeaMicro and AMD spent hundreds of hours in engineering discussions across many weeks and when AMD decided to pull the pin its footwork getting the deal together was “dazzling.” There were reportedly other unidentified contenders for the start-up not all of them chipmakers.

Feldman feels SeaMicro and AMD share the same lust “to change the whole server market.”

Otherwise SeaMicro, which got to this point on $60 million in venture capital, gets access to new markets, resources, technology and scale out of the deal.

AMD said the acquisition doesn’t change its 2012 financial guidance and it expects the transaction to be accretive to earnings sometime after 2012. The deal should close in the next few weeks.

It’s not clear which so-called APU (Accelerate Processing Units) chips started shipping. It could have been either or both the single- or dual-core Ontario and the Zacate. Both are based on AMD’s 40nm Brazos platform and its Bobcat cores.

The Ontario is a 9W low-power model that’s supposed to compete for the heart and soul of the netbook against Intel’s Atom chip. Now, of course, AMD is late in the day trying to squeeze into netbooks, particularly since netbooks aren’t as chi-chi as they were and the Apple iPad has arrived on the scene.

But AMD watcher Nathan Brookwood claims Ontario’s performance could restart the netbook market, making the things more like mid-range notebooks and sensible platforms for Windows and Office with a reasonable five- or six-hour battery life while still at a $300-$400 price point.

Ontario runs too hot for tablets. AMD won’t have a tablet entry for a couple of years when it can get Ontario down to 4W or 5W on a 28nm process with the Ontario and Zacate follow-ons, the two- and four-core Krishna and Wichita. Intel’s also up against it in the tablet market where ARM is the favorite. Now it’s got to worry that Ontario could prove an Atom killer.

The 18W Zacate, on the other hand, could appeal to Apple for its MacBooks or MacBook Airs that currently use Intel’s Core 2 Duos with Nvidia’s programmable graphics chipset. Apple dicks around a lot with graphics acceleration for its own software, Brookwood says. Intel has no programmable graphics so using Zacate could be cheaper and more efficient.

AMD hints that something’s going on between it and Apple. At its annual financial analyst day this week where it was showing off its roadmaps, after a long roadmap drought, AMD happened to flash – without explanation – a slide showing a pair of iMac all-in-one machines, a Mac Pro workstation, Apple’s trademark logo and Fusion branding. Intel gotta pray that it’s flummery, not so much for the volumes but because Apple’s a thought leader.

Anyway, the Zacate, which will compete against Intel’s Celerons, Pentiums and Core i3s, should produce mainstream notebooks and desktops that sell for $400-$600.

Brookwood says the Brazos family is good for a hundred design-wins from serious players that are supposed to launch their wares at the Consumer Electronics Show in January. Intel will be there too with a low-end version of its newfangled APU-like Sandy Bridge chip spec’d at 30W-35W, bigger and hotter than AMD’s widgets with not quite as much graphical oomph but faster evidently. Sandy Bridge is DirectX 10.1-compatible.

JP Morgan Securities checked around and heard that the Brazos chips are great on the graphics end but their “processor functionality is below most AMD and Intel offering.” AMD is very light on benchmarks.

Given the Apple boxes in AMD’s slide it’s probably suggesting that Apple might fancy the upcoming 32nm Llano APU that’s supposed to have two or four x86 cores and a DirectX 11-capable Radeon GPU. It’s meant for $500-$1,000 mainstream notebooks and desktops and should be here the middle of next year. It’s supposed to be capable to 500 gigaFLOPS, Brookwood said. By comparison a six-core Intel Westmere is only good for 100 gigaFLOPS.

AMD also got another little 32nm number called Zambezi for mainstream and high-end desktops that should be out in around next September that has four-eight Bulldozer cores matched with two separate so-called Fusion modules.

On the server side, AMD’s promising to have a chip called Terramar in 2012 with 20 cores for the 2P and 4P market and another one called Sepang with 10 for the 1P and 2P market.

The first would replace a 32nm chip called Interlagos that supposed to have eight, 12 or 16 cores and other would refresh a six- or eight-core chip called Valencia. Both the Interlagos and the Valencia are scheduled for the second half of next year.

All the cores are so-called Bulldozer cores, Bulldozer being AMD’s expected new microarchitecture. And Bulldozer cores are different from what Intel might call cores.

Interlagos and Terramar will follow on AMD’s recently introduced 6000 series and Valencia and Sepang the 4000 series. Neither the 4000 nor the 4000 sold particularly well because they hit the market at the wrong time and being new platforms weren’t drop-ins. They needed new sockets, motherboards, I/O et cetera. The new chips should have that problem.

It all revolves around AMD’s so-called asset light strategy and the fact that circumstances have forced it to spin out its manufacturing plants into a “subsidiary,” otherwise know as the Foundry Company, that will be 62% owned by the oil-rich Persian Gulf sheikdom of Abu Dhabi.

AMD has taken this course to survive but the spin-off raises issues with its age-old x86 license from Intel, which, as it happens, comes up for renewal next year.

The x86 IP covered by the license is Intel’s crown jewels so Intel is understandably touchy about it.

The week before last Intel sent AMD a letter asking for a sit-down to clear up some questions it has about the so-called Foundry Company.

Simply by sending the letter Intel, under its contract with AMD, initiated formal dispute resolution.

Bottom line that means that the only way AMD gets to keep its right to go on making x86 chips is to consign its precious antitrust suit, which has already involved trading more pieces of paper than any other dispute in the history of man, to the ash can – along with its dream of fat juicy award that squares its books.

When the companies sit down, Intel is gonna wanna know from AMD how it is that AMD can unilaterally consign its x86 rights to a third party and what happens to Intel’s IP if and when AMD’s position in the Foundry Company – already reduced from the originally envisioned 44.4% and the deal hasn’t even closed yet – drops yet again.

See, there’s a 20% cap on non-AMD produced wafers in the license, suggesting that AMD would have to own at least 51% of the Foundry.

AMD’s simply calling the Foundry Company a subsidiary isn’t going to cut it for Intel, which wants AMD to make the redacted parts of the cross-license deal public.

Intel is unlikely to get much of an argument about AMD’s lawsuit from Abu Dhabi. The emir could give a tinker’s dam about it. There’s nothing in it for him; he’s just concerned about the validity of the x86 license, which he hopes to leverage into a broader semiconductor-making enterprise. Which is probably why, according to its Foundry deal with AMD, Abu Dhabi can veto any settlement discussions with Intel.

Nobody can do much about the AMD-initiated antitrust case that the European Commission has brought against Intel; that has a life of its own now, but it might be a nice touch if AMD dropped its complaint.

When asked about all this, Intel spokesman Chuck Mulloy said, “Intel does not comment on speculation.” Not to worry, there are just some times it’s best to read the writing on the wall.