Rechercher dans ce blog

Friday, May 14, 2021

Crazy Move #1486: Intel Buys VMware - The Next Platform

proc.indah.link

Does Michael Dell want to be Intel’s largest shareholder? Maybe, just maybe. And there could be an interesting turn of events once VMware is spun off to shareholders in Dell (the company), leaving Dell (the man) as VMware’s largest shareholder, with an approximate 42 percent stake. Imagine, if you will, that Pat Gelsinger, the former head of Intel’s Data Center Group and recently the former head of VMware and now the current chief executive officer at Intel, wants to do virtual CPUs and virtual network ASICs as much as it wants to do physical CPUs and physical switch ASICs.

Imagine, in fact, that Intel decides it wants to buy VMware back from Wall Street, to create a physical and virtual powerhouse.

This is what an intrepid reader of The Next Platform proposed to us and said we should do some game theory simulations on the idea. And once we stopped laughing, we thought, well, why not think about Intel buying VMware? It may end up being more like chaos theory, but here goes.

For the past decade and a half, nothing has curtailed sales of X86 server processors quite like VMware’s ESX Server, GSX Server, and ESXi hypervisors, which came into the market just ahead of the Great Recession, which started in late 2007 and continued through early 2010 or so (in the IT sector at least). Back then, VMware hypervisor running on a single physical server could double or triple utilization by cramming multiple, whole server environments into virtual machines, which allowed many companies to forego a whole generation of server buying at a time when the economy was really messed up and IT workloads were changing like crazy and expanding rapidly. That timing, along with the “Nehalem” Xeon E5500 processors – finally a good X86 server chip from Intel – also launched in 2009, set Intel on the path to dominance in the enterprise datacenter that has been up until now, when that dominance is being called into question by the re-emergence of AMD and the rise of Arm in the datacenter.

Let’s talk about that spinoff for a second. Under the proposal, VMware will distribute a cash dividend of $11.5 billion to $12 billion to VMware shareholders, of which Dell Technologies is the largest, with an 81 percent stake. The rest was sold off to the public by EMC, which technically should have been illegal as far as we are concerned. Why? Because when EMC bought VMware in 2004, EMC shareholders already owned all of VMware and it should have been illegal to sell the initial 15 percent stake of VMware separately to Wall Street, which grew to 19 percent over the years. This was always a bit too much like Max Bialystock from The Producers for our taste. (You can’t sell 25,000 percent of anything, even if it isn’t a flop. Or as one of our mentors used to say, “Anything worth selling once is worth selling two or even three times. . . .”) With the spinoff, Dell (the company) gets $12 billion or so, and $9.7 billion of that goes to Dell Technologies. In exchange, the individual Dell shareholders plus the existing VMware shareholders take control of VMware, which is a separate company already trading on Wall Street separately, and a part of the $67 billion EMC-VMware acquisition that Dell did back in October 2015 unwinds.

When this is done, Michael Dell will control 42 percent of VMware personally and his long time cash line provider, private equity firm Silver Lake Partners, will control 11 percent. This means that as long as Silver Lake agrees with Michael Dell, Michael Dell really runs VMware, not Raghu Raghuram, a long-time top executive from VMware who we have known forever and who was just appointed chief executive officer, replacing Gelsinger. Dell, the man, may be happy to have Raghuram actually run VMware, as he was happy to let Gelsinger do it before him. But make no mistake about who is in control.

Now, if Michael Dell really wanted to up his game and get in on the ground floor of a company in transition, and possibly help Dell, the company, as well, he might approach Gelsinger and propose that Intel take over VMware. And with Nvidia spending $40 billion (mostly in stock) to buy Arm Holdings and with AMD spending $35 billion (mostly in stock) to buy Xilinx, we think there is probably a way for Intel to acquire VMware mostly in stock and not use a lot of cash it needs to ramp up its Intel Foundry Services business to take on Taiwan Semiconductor Manufacturing Corp and Samsung Electronics in advanced semiconductor manufacturing, as is the plan Gelsinger has come up with along with creating better chips for clients and servers.

As we go to press, Intel has a market capitalization of $216.5 billion and has issued just over 4 billion shares on Wall Street. As it exited the first quarter, Intel had a trailing twelve month revenue of $77.7 billion, $20.3 billion in operating profit, and $18.7 billion in net income. As for Intel’s balance sheet, the important things are $5.2 billion in cash, $2.4 billion in short-term assets, and $14.8 billion in trading assets that can be liquidated as need be. Ditto for Intel’s additional $5.4 billion in equity investments and $1.4 billion in other investments. Add that up, and it is $29.2 billion. Intel has promised to spend $20 billion on foundry expansion and stop its share buyback nonsense, and it keeps generating more than $4.5 billion a quarter beyond that. So it can blow a lot of its cash because the coffers will refill for a while – so long as the IT market is expanding.

VMware has 77.7 million shares on Wall Street today and a market capitalization of $65.6 billion, which would make an acquisition very expensive. More expensive than what Michael Dell did to do the EMC-VMware deal, the biggest one in IT history. Intel would have to massively dilute its shares to cover the bulk of a VMware acquisition, but the good news is that there is very little chance another buyer would emerge. But clearly, to take control of VMware requires the consent of Michael Dell and Silver Lake because they own more than 51 percent as a pair.

In its trailing twelve months (which runs through the end of October 2020 because VMware has not yet reported the end of its fiscal 2021 year), VMware has $12.1 billion in revenues, operating income of $1.7 billion, and net income of $1.6 billion. At this point, it is a steady freddy, legacy systems software platform with over 300,000 customers in the enterprise datacenters of the world. In 2013, there were 36 million VM guests on the ESXi hypervisors of the world, and we estimated in 2018 that had nearly doubled to around 60 million guests. It is probably closer to 90 million guests these days. VMware has not reported numbers for the vSAN virtual storage and NSX virtual networking bases in a long time, but vSAN could be in the range of 30,000 customers these days and NSX could be in the range of 10,000. Those are pretty big bases, even if only a fraction of the VMware ESXi installed base.

What VMware only has left is just under $4 billion in cash as it has already paid out a huge wonking dividend to shareholders, as you can see in the chart above. Cash in the acquired company is always part of the equation, and in this case, it doesn’t change the price of the deal all that much.

The VMware business would snap right into Intel’s Data Center Group, which may be particularly important if many of the hyperscalers, who represent on the order of 35 percent to 40 percent of CPU sales for Intel in any given quarter, start designing chips and having them fabbed for their own private use, as Amazon Web Services does for CPU and DPU chips today as well as for AI training and inference. Something has to fill that revenue and profit gap, which we think is going to open up, and doubling down on enterprise customers who will absolutely pay a lot for legacy platforms so they don’t have to change them. Enterprises like to evolve things very slowly, achingly slowly over decades, until they can’t stand it any more than then they try to throw it out – and a lot of the times, they still can’t. IT is not their business, as it most certainly is for hyperscalers and cloud builders.

In the trailing twelve months, Intel’s Data Center Group had $24.7 billion in sales and $8.4 billion in operating profit and that operating profit is coming down and it is going to keep coming down as AMD competes hard in servers and the Arm collective (represented by Ampere Computing and those who are building their own Arm server chips like AWS) keep hammering on Intel. In fact, VMware operating profit is rising and converging on where Intel’s Data Center Group profit is heading, so blending them somewhere around 2022 would not hurt either company. And, after Intel sells off the NAND flash business to SK Hynix (no, you do not get to randomly uncapitalize a proper noun, sorry) for $9 billion, the combination of Data Center Group and VMware would create a bigger datacenter business than a client business at Intel. If that matters.

We think it might matter to Gelsinger, who is looking for wins, and who knows VMware better than he does? Who knows better the competitive issues it faces? Or the beauty of a legacy platform that can be milked?

The question is, does Michael Dell appreciate this, and does he want to be the largest shareholder of Intel but lose his controlling stake in VMware as part of the bargain. Just to pay for VMware without as premium, Intel would have to issue 1.2 billion shares, which is a lot. The premium might have to come from cash, and Michael Dell would get his 42 percent of that to put into his own pocket, so that is a motivation. But his stake in Intel, unless he already owns a lot of it privately, would only be around 10 percent. Still, that is probably a lot of sway separate from the control Dell, the man, has over the largest OEM server and PC supplier in the world.

Intel could certainly use VMware to do a better job attacking virtual storage and virtual networks, and to come up with a coherent and enterprise-friendly DPU story.

Intel buying VMware once it is set free is an interesting idea, and not the stupidest thing we ever heard, considering all of the chess pieces on the board. For all we know, this is the plan that is cooking. And if not, you can bet Dell and Gelsinger are thinking about it right now.

One final thing. Here is the secret decoder ring to decipher the title. The following are the architectural generations of X86 processors relevant to the datacenter, subject to debate we realize. So this crazy idea ranking is a bit of an Easter egg. We are only counting architecture “tocks” and not process shrink “ticks” as CPU generations.

  1. 8086: The beginning of it all, also known as the iAPX 86. (1979) The 8088, launched in the same year with a skinnier I/O bus, is a variant of this, and it is arguable that it is really that different by modern standards.
  2. 80186: Also known as the iAPX 186. (1982)
  3. 80286: Also known as the iAPX 286. (1982)
  4. 80386: Renamed the i386 because, you know, marketing people gotta do something. (1985)
  5. 80486: Officially named the i486, but no one cared and called it the 80486 anyway. (1989)
  6. 80586: Officially the Pentium 5 or P5, and bringing you the architecture of the first credible servers based on the X86 architecture. (1993)
  7. 80686: Pentium Pro (1995) and Pentium II (1997), called the P6 architecture internally.
  8. 80786: NetBurst or P68 architecture, used in Pentium 4 chips for desktops and servers. (2000)
  9. 80886: Core architecture (2006), the laptop saves the server day with the “Nehalem” Xeon E5500 (2009)
  10. 80986: Sandy Bridge architecture (2011)
  11. 81086: Haswell architecture (2013)
  12. 81186: Skylake architecture (2015)
  13. 81286: Ice Lake architecture (early 2021). Yes, that is a mighty big gap, but not as big as the one Intel left between the 32-bit Xeons and the 64-bit Itaniums that AMD drove a Sledgehammer, a Bulldozer, and a Piledriver through with the Opterons. The hole Intel left this time around with the Xeon SPs was Epyc enough, though, to point out that the Intel emperor was wearing a tankini like Borat. (No, you can’t unsee that image.)
  14. 81386: Sapphire Rapids architecture (late 2021)
  15. 81486: VMware represents a new architectural layer under Intel’s control and on top of X86. Maybe we call it 81ESX? (2022?)

Thanks for playing.

The Link Lonk


May 14, 2021 at 12:49AM
https://ift.tt/3bq6kfD

Crazy Move #1486: Intel Buys VMware - The Next Platform

https://ift.tt/2YXg8Ic
Intel

No comments:

Post a Comment

Featured Post

Intel Falls on Latest Server Chip Delay; Rival AMD Gains - Yahoo Finance

proc.indah.link (Bloomberg) -- Intel Corp. fell after saying a new version of its Xeon server chip line will go into production in 2022, r...

Popular Posts