Contact is head of financial services vertical at $Amazon (A...

Contact is head of financial services vertical at $Amazon (AMZN.US)$ AWS. Probably worth remembering that senior execs operate at very high levels and typically lean optimistic - so need to stay sober but was pretty damn bullish.

Punchlines are that:
-1) Fin svcs is the largest vertical in AWS (25b) and will accelerate from 15% last year to 25-30% this year
-2) Everyone is working on moving 100% of their data to the cloud to deploy Gen AI across their organizations
-3) Margins in AI platform services are considerably higher than in just consumption
-4) $NVIDIA (NVDA.US)$ GPUs in very high demand (“The GPUs. We just don’t have enough of them.”).

Was also surprisingly positive on $Snowflake (SNOW.US)$ ( $Amazon (AMZN.US)$ is their closest partner), but recommends customers who use multiple clouds go with $Databricks.

Highlights
-Runs financial services vertical within AWS – is 25b out of 80-85b in AWS [AWS did 91b last year so clearly not right on top of numbers]

-People now understanding that building AI models across the organization has a considerably larger ROI than single use cases. And only way to do this is by moving all data to the cloud.

-Typical first deployments for AI have to meet 25% ROI threshold bc otherwise not going to get accepted. Easy to do for 3 best use cases: contact center, content compression and monitoring.

-Fin svcs is typically fastest growing vertical and also consumes the most AI / GenAI use cases which are 50% of all use cases for the vertical

-Largest accounts have 25% of data in the cloud and were planning on going to 50%. But in Gen AI world, want 100% in the cloud to drive models. CapitalOne reformulated their strategy and said will migrate 99% of their data in order to deploy GenAI across applications. NatWest in UK doing the same thing.

-JPM committed to a 10b project to put all data in the cloud, and AWS is primary vendor.

-Regional banks are a 60b opportunity. Talk w/ them weekly and they are under a lot of pressure to be more efficient – see CEOs getting involved in Gen AI discussions.

-Celent is IT consultant. Did survey of F1000 and IT budget expected to grow 7% this year. Haen’t seen that in forever – typically IT growth expectations never tops 4-5%. So that extra 2-3% will create much higher growth in cloud.

-Think fin svcs AWS will grow 25-30% this year barring a financial crisis up from 15% last year. Most of acceleration will come from internal AI use cases which are less regulated – later on will figure out consumer-facing regulation that is a constraint for now.

-Also seeing acceleration in other verticals where AI is viewed as critical: auto industry w/ real-time diagnostics. Have signed 2 huge deals w/ automakers in last three months for this reason

-Margins in AI offerings are considerably higher than in just consumption. Hardware is lowest-margin product – but databases, services, solutions, etc. have much better margins for AWS.

NVDA Highlights
-If $NVDA doubles production this year as they said, still have a problem bc we need at least 10x more to satisfy all use cases. So this is going to be a 3-5 year marathon rather than a sprint.

-60% of our budget is going to be refactoring data centers for GenAI

-“AWS has the largest private network in the world. And that private network is not going to be useful in the world of GenAI. So we are cosponsoring with NVIDIA project to reimagine the network of the future” [I’m not sure I understand this point]

-The problem w/ GPUs is they don’t talk to each other very well. Took us a long time to get $AMZN chips to talk to $NVDA.

-What other limiting factors? “The GPU’s. We just don’t have enough of them. There is no question that in three years, most of our data centers are going to be at least 50% GPUs.”

-“Well, you see, again, we want to buy as much as NVIDIA is willing to sell. The key point here is NVIDIA has other verticals that they have to serve in addition to us. They have automobile. They have the gaming industry. They have the media and telecom industry that they also serve. And so we are in a pecking order, and we don't have much choice about how many chips we get.”

Other Highlights
-Think $Snowflake (SNOW.US)$ is well positioned for GenAI bc 1) can collect images, natural language, photos, audio which is important bc lot of GenAI models are going to be augmented w/ social media; 2) want to build a data lake and SNOW sells both he software but the data strategy solution; 3) proven methodology for consolidating on-prem databases – will give you the roadmap for 100 databases and do it for you.

- $Snowflake (SNOW.US)$ is mix of db provider and system integrator.

-Think $MongoDB (MDB.US)$ db is as good as $Snowflake (SNOW.US)$ [now losing credibility as these are two totally different technologies]. But they don’t have the solution services like $SNOW

-$Databricks pretty good but missing service component. Any customer using multiple clouds for data should go with Databricks over $SNOW

- $Accenture (ACN.US)$ , $Wipro (WIT.US)$ , $Tata all coming to us more frequently bc they don’t have enough business. Think that so much work that SIs did is getting automated. Think a lot of things that Wipro provides are going to disappear in the future.