Supply chain agility - an achievable goal? How NVIDIA uses SAP Business Network and AI-enabled planning to minimize the gotchas
- In this SAP Sapphire Orlando exclusive, we share the story of NVIDIA's push for supply chain excellence - and how the SAP Business Network fits in. Cloud-based planning with SAP IBP also plays an important role.
Credit where due - SAP had no shortage of customers to interview at SAP Sapphire Orlando. It wasn't easy to come up with a shortlist - but NVIDIA was on it.
Why? Because NVIDIA is knee deep in solving supply chain problems that flummox organizations. Plus: I knew NVIDIA had an SAP Business Network story - something I haven't covered in prior SAP pieces.
NVIDIA's Ron McCutchen, VP of IT Business Applications & Engineering, was in Orlando to deliver his session, "Build Resiliency Across Your Global Supply Chain with SAP Business Network." The event gods aren't always kind; I was doubled-booked during his presentation. So I asked McCutchen: what did I miss?
NVIDIA's business network story - fueled by the Mellanox acquisition
NVIDIA's business networks story picks up steam after their $6 billion acquisition of Mellanox in 2019. NVIDIA's ambitious move into "high-performance interconnect technology" expanded their scope - and their challenges. The Mellanox acquisition gave NVIDIA the confidence to state:
With Mellanox, NVIDIA will optimize datacenter-scale workloads across the entire computing, networking and storage stack to achieve higher performance, greater utilization and lower operating cost for customers.
At the time, NVIDIA's CEO said that the emergence of AI and data science fueled "skyrocketing demand on the world's data centers." He added:
Addressing this demand will require holistic architectures that connect vast numbers of fast computing nodes over intelligent networking fabrics to form a giant datacenter-scale compute engine.
But to get there also requires an optimal supply chain architecture - and that's where McCutchen's team comes in. As he told me in Orlando:
A lot of people think of our GPU and the graphics and gaming and all that, but the company really has moved way beyond that last five years to full stack, all the way through software, to the hardware platforms... So we acquired Mellanox two years ago; we've now fully integrate them into our ecosystem.
We've been using the SAP Business Network for our contract manufacturers. It's a fabulous environment. We leverage the network with different companies for the making of the wafers, the assembly test, the packaging - all those types of things. So we have a very connected global network that we use.
SAP Business Networks - another go-live on deck
The business network allows us to connect into all of those suppliers and vendors. With Mellanox, you have some different contract manufacturers and suppliers. Part of the integration process - and our go-live here at the end of this month - is onboarding those vendors onto the network so we could leverage them. So for us, why we use it overall is really speed and agility.
I'm not going to knock speed and agility. But we've heard plenty of that from the keynote stage. How does that result happen? McCutchen explains:
The business network provides a solid platform for us that we don't have to build ourselves; we don't have to customize. It comes with standard adapters; it comes with a multitude of ways to integrate. Your vendors are not all the same. Some are using XML; some are using files; some are using real-time integration. So we've really leveraged that to be able to integrate Mellanox.
For the Mellanox integration, this supply network is required. But it's more than that. In today's constrained/volatile environment, a different level of agility must be achieved:
The broader picture is we really use it to manage our network of all our global suppliers and vendors. It gives us a great deal of speed and agility. We're in a very constrained environment from a supply perspective. We need the ability to quickly be able to expand with an existing partner or onboard another partner in weeks, to be able to support the changes - based on the very constrained environment that we're in.
This isn't a pandemic economy disruption anymore. Upheavals will be a constant; legacy approaches will fail. McCutchen agrees:
Absolutely. Every month, every quarter, something new comes up. Suez Canal gets shut down, a new tariff or regulation; a new export control happens. Or some other geopolitical event... a city the size of Shanghai gets shut down, or some other massively disruptive event. We have to manage to the best that we can.
Have NVIDIA's network investments made a difference here?
Our speed to market can be weeks, not months, to be able to onboard partners on the network. Many of them are already on the network. Many of them are familiar with the network. Having the base platform there for you, with the options to integrate, is a huge advantage. We're not recreating the wheel each time; we're leveraging something existing, in many cases, that partners already know. So that helps tremendously.
On SAP IBP, open platforms, and the single source of truth
Then the interview took a twist. McCutchen dropped an intriguing line about his other SAP Sapphire presentation: co-innovating on AI/ML with SAP IBP (Integrated Business Planning). How did that come about? Before McCutchen's team moved to IBP, it was that classic planning story: Excel files everywhere. McCutchen:
A lot of our planning processes involve Excel and a variety of other tools. So we had all the common problems that have been around forever: single source of truth [issues], latency in the data, redundancy of information, people massaging information, 'My numbers don't agree with your numbers' - you name it.
So just moving to IBP to start with, to consolidate a single data source for demand planning, was huge... I've been working with SAP products for 30 years. Two things SAP has not been known for is a great user interface, and open platforms. With IBP, you get that.
Even better: planners don't have to give up their attachment to their trusty spreadsheets.
The Excel data integrates right back into IBP. So it's still the still the same single data source, and the planners love Excel, so they can still do it in Excel.
A cloud-based planning system is a huge plus for McCutchen's team:
They didn't just take APO and stick it in the cloud. It's a rewrite from the bottom up - it's cloud native. So you get all the advantages of a cloud-based application. The quarterly releases, enhancements, all those things come right out of the box, the performance you get and all the rest of it.
And where does AI fit in?
We have all kinds of demand signals that come into our channel, buying habits and different things we use to sense demand. And then on top of that, we can go back actually ten years, and look at certain things - recessions, economic slowdowns. The buying habits repeat themselves. IBP really can't absorb the volume of data that we have over such a long time period, and be able to run AI and ML against that.
So we use our own GPU and AI and ML technology against some of those demand signals. We eventually then loaded those at a kind of a higher level, into IBP, to help some of the demand signals on the planning. We run our statistical models and the algorithms, and all that good stuff.
Has this additional data paid off?
Yes - we got better visibility into the channel and predictive models. Just being able to run the what-if scenarios that feed into our request build plan, as we call it, has been tremendously helpful. It will get even better over time.
Sounds nifty, but I need proof points. McCutchen had a good one: by applying these demand signals, McCutchen's team was able to largely anticipate the surge in NVIDIA equipment demand caused by the crypto craze - not too shabby. McCutchen:
I think we've been able to pretty well control it, even anticipate at the macro level, there's always kind of smaller adjustments to that. Nobody wants to surprise the CEO with a 'We screwed up, and we totally blew that demand.' That's what we're trying to avoid.
I'm guessing other SAP customers were very interested in this presentation?
We got a lot of questions afterwards.
The wrap - NVIDIA, SAP, and... the metaverse?
I was struck by the impact of gaming culture on NVIDIA. As much as I poke fun at the metaverse as a future framework for the enterprise, I respect how gamers push tech to its limits. McCutchen told me that keeping up with gamers changed NVIDIA's culture for the better:
Building an architecture which is more agile and flexible to change, and not highly customized, you're right - gaming forces us to do that. I think that pays off in other parts of our business like data center, which is now really our largest business group.
Looking ahead, is there anything McCutchen needs from SAP? That question brought an unexpected answer: the enterprise metaverse.
We've had a good partnership with SAP. We're looking at a number of co-innovation opportunities with them. Some exciting things with S/4, and potentially our Omniverse. The release of our CPU Grace, which is ARM-based. There's opportunities for SAP to run on ARM; there's opportunities for us to go to market with our Omniverse, and not only the planning tools but supply chain...
As NVIDIA moves core systems to S/4, the Omniverse could play a role. I told SAP, 'We're going to go build our whole supply chain in Omniverse first and optimize it, and then we'll go build out our configuration processes in S/4.' A lot of companies would benefit from doing that, being able to model everything first, optimize it, and then go build and configure it in SAP.
I bash the metaverse hype for lacking enterprise use cases. However, what McCutchen is talking about here is an enterprise use case for sure, perhaps an evolution of the digital twins concept - one of the few so-called "metaverse" scenarios that already has enterprise credibility. Another one to watch.