LLMs will be stressed by enterprise systems, says Wedbush's Sherlund

Melissa Lee (Host) 00:00.070

our next guest expects the AI trade to shift from broad enthusiasm to a more selective environment in the new year wall street legend rick sherland is founder of sherland partners and recently joined wedbush investment banking as a senior advisor rick great to have you with us

Melissa Lee (Host) 00:12.590

nice to see you

Rick Sherlund (Founder) 00:13.910

melissa thanks for having me and i should caution you that i no longer wear the analyst at so i'm going to be ducking some stock questions probably

Melissa Lee (Host) 00:22.190

that is fine we we want to get into your head rick in terms of how you're thinking about this AI trade so when you say shift from broad enthusiasm to being more selective what do you mean and and you know what we saw today in terms of the enthusiasm behind the hardware names

Melissa Lee (Host) 00:36.270

versus some of the others you know does that play into your thinking

Rick Sherlund (Founder) 00:40.950

i think this year has been one of kind of a monolithic trade people think of the max seven it's now about thirty five percent of the S and P but i think twenty twenty six you're going to see that trade broaden out very substantially you've got a whole universe of SAS companies

Rick Sherlund (Founder) 00:57.710

that the the streets kind of looking beyond right now they're some issues with respect to is the architectural architecture flexible enough to accommodate AI is the business model at risk so they've been favor but i do think you're going to see some of those come back where

Rick Sherlund (Founder) 01:13.910

they've made good progress on both sides of adopting AI and the business model changes you'll see a very robust IPO market because there are a lot of the private AI companies that have gotten to scale and some of the SAS companies that didn't get out in the peak of the market in

Rick Sherlund (Founder) 01:30.390

twenty twenty one will likely be coming coming to market where they've made good traction on AI and the business model shifts M and A of course is going to be a big thing next year all the big enterprise companies need to add AI into their own architecture and an easy way to do

Rick Sherlund (Founder) 01:48.750

that is through acquisitions so i i think that all the derivative beneficiaries of AI in the streets and you guys have been looking at pretty closely i think there's a lot more of that that we'll see next year not just within software but broadly across the economy so it'll go

Rick Sherlund (Founder) 02:06.800

ahead yeah

? (Analyst) 02:07.680

sorry rick tried to cut you off because that's exactly where i wanted to go and to better understand your thoughts and i'm not gonna ask you to pick stocks here but when you talk about enterprise adoption and really really the follow through from the are we talking though about

? (Analyst) 02:20.950

the industrial sector and people finally or you know banks places that have had a great run industrials have rallied all year i think on some level because they were already running their business more efficiently are you really talking though about the platforms a little closer

? (Analyst) 02:32.670

to home especially whether you are talking about the hardware names i certainly agree that software has underperformed here as people have been concerned about the hardware and i think you'll get follow through but please just clarify this this enterprise adoption dynamic sector

? (Analyst) 02:45.470

wise if you can

Rick Sherlund (Founder) 02:46.470

yeah so a couple thoughts you know AI has been a novelty for consumers and with chats and a lot of knowledge workers have gotten big benefit out of it because they're very good at researching and summarizing content for for for knowledge workers but after three years on the

Rick Sherlund (Founder) 03:04.230

market now it's finally becoming part of business process and and workflows how a company actually conducts business so this involves the use of agents or agentics and how they collaborate and do complex workflows as a result the amount of inference that's going to be required

Rick Sherlund (Founder) 03:22.510

is going to explode you know a single call to a chatbot means it's one call to the LLM to give you the answer to to for inference if it's a workflow related question particularly for complex workflows it could be ten or fifty trips back to the LLM so LLMS are going to really be

Rick Sherlund (Founder) 03:42.710

stressed by enterprise systems we've got reasoning we're asking these models to do now they're not built for reasoning so they just kind of throw a lot of weight into this GPU saying give me a hundred answers and i'll pick the best lawn so i think for all these reasons you know

Rick Sherlund (Founder) 04:00.390

the the inference is going to be the heartbeat of american well global business it's how it runs so that's going to create enormous demand for data centers so the data center trade you know i'm not worried about it we're going to have so much more demand than we have supply of

Rick Sherlund (Founder) 04:18.110

data centers given all the constraints on data centers the capital the GP us memory etcetera so the question would be and i think this has been forefront in people 's minds is well is the business model there to support the financial commitments that are being made and i think

Rick Sherlund (Founder) 04:36.270

people look to the LLM market and this the leadership in LM changes over more times than a you know a close match basketball game you know the score just keeps going back and forth so the chinese have open source models you've got meta and and nvidia with open source models so

Rick Sherlund (Founder) 04:54.150

the LLM business is going to be a very competitive business but i don't see the leaders in that market saying we're going to make all our money on LLMS what they're going to do is what microsoft did with windows and go up the stack they're going to do what oracle did from the

Rick Sherlund (Founder) 05:09.110

database and move up the stack