We’re still waiting for movement by someone on DoJ’s opposition to the HPE/Juniper deal. Meanwhile, the two companies, their customers, and their competitors are all gaming out the results of the possible outcomes. That’s difficult because we can’t be sure why HPE or Juniper wanted the deal in the first place.
The general view in the market, and on Wall Street, is that HPE wanted Juniper’s AI. I’m more skeptical of this as time goes by. If Mist AI was the target and if DoJ is worried about the impact of the deal on the enterprise WiFi market, why would HPE not offer to divest that part of Aruba? In any case, it seems to me that buying Juniper to get Mist would be like buying a dozen eggs (at their inflated prices) to get some egg cartons.
I think that the Juniper deal is more likely to be HPE’s reaction to two forces. One is the force of commoditization. Everything in enterprise IT has been under price pressure, and that’s not going to stop now. In a commodity market, you need to maintain account control and sell as many pieces of a project as possible, or you’re locked in a price war. The other force is IBM. Here’s a company who is constantly underrated by the media, but their stock has been up in a down quarter, and they have the most strategic influence of any company according to my data. They also handle almost three-quarters of the commercial transaction processing globally, and of course the ownership of transaction processing means owning the data and also owning what drives networking.
IBM doesn’t sell network gear any longer, so if we view computing, storage, and networking all collapsing into a commoditization black hole, then you can’t afford to have a missing piece. Juniper would give HPE a big footprint in a space where IBM has only nail clippings.
For Juniper, you have to ask why, if Mist AI is an asset worth buying a whole company over, they can’t overwhelm their traditional rival, Cisco, with it. The fact is that a commodity market doesn’t easily admit to feature differentiation at all, and network equipment has always been a case of Cisco versus not-Cisco, meaning that Juniper doesn’t have a realistic chance of gaining market share on Cisco, only gaining share in those who reject Cisco out of hand. The problem Juniper has is that it’s a lot easier for Cisco to gain not-Cisco market share, and a good way of doing that would be through increased strategic influence. For several decades, the real driver of enterprise network success has been the data center network, and data center network decisions are driven by the compute side, which is why IBM has so much influence. Cisco has at least some server position, but Juniper doesn’t. HPE, of course, does, and is second to IBM in influence there.
So, let’s assume all this is valid thinking. How do we game the possible outcomes?
OK, DoJ drops its objections, with or without concessions from HPE, or they lose the case. In this situation, HPE takes a big forward leap in terms of market importance. They don’t need to architecturally integrate Juniper; they’re added products for HPE to leverage in deals they control at the strategic level. They now have something IBM does not, and they’ll surely threaten IBM with it.
Who loses? IBM in the long term, because they will either have to buy a network player or use their influence to turn data center networks into a white-box lake, with no dominant strategic player. In the short term, Cisco, because HPE/IBM battles shift focus to something Cisco isn’t much of a player in, which is the compute space.
This might create another winner, Broadcom. Obviously, if IBM needs white-box supremacy in the data center, Broadcom chips win out. Similarly, any battle for influence that elevates the data center will elevate VMware. In fact, Broadcom might be a winner no matter how the deal works out, as I’ll get to.
Now for that outcome. The merger is not approved, so what happens? HPE and Juniper both lose. The former now has less to fend off IBM from one side, and Dell and Broadcom/VMware from the other. They now have to make Aruba into a decided asset, which diverts them from the defense of the server space. The latter is now facing Cisco, IBM, and HPE/Aruba in a battle for influence, and that will be very difficult for them to win, or even bring to a draw.
Who wins? Cisco and IBM, obviously, but also Broadcom. Platform software like VMware’s stuff is part of the server/compute picture, and a strategic battle at that level helps them whether the battle is offensive or defensive. Not only that, I think that the future of networking really lies in virtual networks, and VMware has a long-standing NSX positioning there. Both Cisco and Juniper have virtual-network stuff, but both companies have been cautious lest they undermine the feature credibility of the real hardware they sell. Broadcom has no such concern.
What about enterprise AI? How might that be impacted (if at all) by the success or failure of the merger? I think that AI becomes more important if the deal goes through, because it magnifies the compute battle, a battle IBM knows is really about AI as part of the transaction workflow rather than AI as a personal copilot. If the deal fails, I don’t see HPE or Juniper driving enterprise AI any differently.
I think the deal should go through, not for reasons of market optimality but as a matter of law. I don’t think DoJ has presented a valid reason to stop the deal; the notion that somehow enterprise WiFi is threatened with monopoly is, to me, simply silly. It would be a shame to do something market-negative for a silly reason, but politics is politics after all, and silliness is not unknown there.
