Hi team, I wanted to share some thoughts, I’ve been pondering over this for a while, but been stupidly lazy to sit and write this down. But got some time now so here goes:
Before that, Disc: took a position in Infosys recently.
First thoughts, companies like meta, microsoft, google, and amazon were traditionally seen as asset light firms, and that was their moat. But of late over the last few years - all that free cash flow generation has been going into building data centre - hyperscalers
So that ease of mobility isn’t the same. With these investments - they’re going to be having payback periods but the GPUs being used for these hyperscalers are depreciating very fast, so at some point this large scale of investment has the risk of being obsolete. If one were to follow Moore’s law (transistors in the chips keep doubling) then eventually higher computing power will be available for cheaper prices sooner or later, and at that poitn data centre payback periods may increase.
Currently capital is easily available from private equity - investors like what they’re seeing so firms like blackstone etc find it easier to raise capital for such “AI” projects. This AI boom is mainly funded by equity privately owned.
Hence that risk may get tranferred when Open AI and Anthropic comes out with their IPOs.
The risk here is not complete destruction of value but rather - just a small slowdown or change in expectations can take the valuations from the sky down to earth. More possible when equity is listed as compared to private markets.
Case in point: please watch Mark Cuban’s argument that OpenAI will never be able to pay back the trillion they’re raising. https://www.youtube.com/watch?v=oEVHNvE_jDw
But yes - if one believes in the growth of AI - then they must also equally believe that what you see today isn’t the final stage - and there is more yet to come. In such a scenario - it would be unwise to bet on a single LLM or a single chipmaker without knowing the future. Too much innovation related risk.
And this is what Indian IT Firms have done actually! Infosys has Topaz, TCS has TCS WisdomNext. Basically indian firms have tried to formalise the AI integration process - using multiple different AI models - so as and when a better tool is available that can be easily integrated through this platform. yes indian firms are trying to stay relevant. TCS is the only big Indian IT firm which has set aside funds for investments in Data centres (project called HyperVault) over next 5/7 years, this will be asset heavy and will dilute their above average return ratios.
(If you watched the Mark Cuban interview, you’d notice he praises Apple - they have gadgets that can adopt the winner of the AI race. They’re uniquely positioned for this. Indian IT firms can also be thought of in a very similar manner, it’s not gadgets but rather enterprise integration expertise, that’ can employ the winner of the AI race.)
But pureplay AI firms already have risk-on from their capex heavy investments, and they would rather outsource the integration part to service firms. I think Indian IT has taken the right decision, and I’m glad. (I stick to this despite the small intiatives from Anthropic and OpenAI to set up integration from their end for enterprises).
(Another example: Anthropic had created an AI based HRMS, but Workday CEO claimed that despite having created this, Anthropic with its small team still used workday as its HRMS ironically)
Social media points to Mr Murthy and other leaders not being innovative enough but it’s not a lack of risk - but they’ve done a good job protecting shareholder money without taking a risk into the unknown. For that kind of risk - they have their own VC firms like catamaran, etc.
So doing buybacks, and slowly acquiring niche firms is the way to navigate unknown territory whilst also keeping yourself relevant to clients by providing AI integration tools. Legacy systems is the key word here because there is so much complex minute data for all these businesses that it’s really hard to immediately put on to an LLM and expect automatic integration - that will cause more harm than benefit. So this is a long drawn process, and reimaging how this can be done will generate more revenue streams for service providers.
A sepculative guess here - by partnering directly with AI only firms - you also the run the risk of sharing all your data, and them learning from private data, which many firms would rather avoid, hence bringing a trustworthy IT firm - with long standing domain expertise and relationships is key.
Another segment that is compeltely being ignored is the Engineering R&D department. Indian IT firms are split into IT services and Engineering R&D firms. For example LTI Mindtree is entirely IT service based but LTTS is a pure play E R&D firm. Amongst the big players, HCL, TCS, Wipro and Infy have sizeable E R&D department in this descending order as of 2023.
Considering what I know about E R&D, it involves physical and digital. You are talking about supporting firms to develop their patenta in medical, oil/gas, automation across the manufacturing space etc. You help them design, and verify at multiple stages. This segment is not going to be affected by AI in the literal sense as we are seeing right now. It will be augmented by AI, but not displaced by AI.
A key aspect to note is what acqusitions are IT firms doing at the moment:
Wirpo bought Harman DTS - a key E R&D player based in US.
HCL Tech acquired - ASAP group specialising in e-mobility
Infosys has purchased the most in the e R&D space ramping up their focus over here.
Kaleidoscope - firm specialising in medical devices and other highly specialsied industrial gadgets (worked with canon medical, P&G etc)
InSemi - contrary to the notion that no indian it firm acquires Indian startups - here is some change. InSemi is an indian semiconductor chip design firm.
InTech - A german firm sepcialising in software and electronics e r&d for Automotive, Railway and Smart Industries (e-mobility and autonomous driving)
Mr Pareekh Jain founder of EIIRTrend, provides insights into what’s happening in the IT industry, and I found this data to be useful.
Reserve the right to be wrong :)
