IGI Logo

Baking in higher Gen AI disruption risks

We bake in higher Gen AI-driven revenue deflation in FY2027-28E, noting increasing risks even as the current adoption rates align with our expectations. We expect the weak transmission of tech spending growth to services to continue leading to moderate industry growth in the next several years. We bake in higher disruption risks in the future by increasing the cost of equity assumption by 50-100 bps. These together drive cuts in fair values for companies under our coverage. We expect IT services to remain relevant in the long term and do not change our terminal growth assumptions. Quality challengers can have a leg up over incumbents. Prefer Infosys, TCS and TechM among Tier 1 names, with Coforge and Hexaware among mid-tier. Upgrade the rating of PSYS to REDUCE from SELL.

Bake in higher Gen AI-driven revenue deflation in the next couple of years

We bake in ~3-3.5% revenue deflation for the IT services industry in FY2027-28E. Note that we expected a 2-3% deflation earlier. The faster pace of innovation, focus on automating software development by key AI labs, high rate of adoption by the developer community and AI-first mindset of enterprises make the upper end of the deflation range more likely, in our view. We slightly tweak Gen AI efficiency assumptions (Exhibit 1). The higher deflation assumptions drive ~1-2% cut in US$ revenues for FY2027-28E.

We assume weak transmission of tech spending growth to services to continue

Indian IT is an incumbent—we expect growth similar to the industry. The global IT services industry grew at 5%, on average, in the past 10 years (Exhibit 2). However, global IT services growth has trailed technology spending growth and long-term average growth in services in the last couple of years (Exhibit 3), a trend we expect to continue. As a result, while we expect robust overall technology spending growth in the next few years, driven by increased Gen AI adoption, we expect a larger portion of the benefit to be captured by frontier labs and hyperscalers, with a far smaller component flowing through to services. Better macro and lower Gen AI deflation headwinds (post FY2029) can drive better services growth. We bake in Indian IT industry growth to be similar to the overall industry growth for the next decade, in the 4-5% range.

Target multiple of ~13-18X and FV cut of 15-21% for Tier 1 IT

The reduction in revenue growth estimates leads to a 1-3% cut in our EPS estimates across companies. We arrive at the PE multiples for stocks using a 2-stage DDM. We lower our growth rate assumption in the high growth phase to 5% in US$ terms for Tier 1 IT. We maintain terminal growth of 5% in rupee terms (3% US$ growth + 2% rupee depreciation) for all stocks. We bake in the risks from possible higher AI disruption and continued macro pressures by increasing our cost of equity assumptions by 50-100 bps. These changes result in a target multiple of ~13-18X for Tier 1 IT and 18-27X for mid-tier on March 2028E earnings. These together drive a 15-28% cut to FVs across companies. These are shown in Exhibits 4-6.

Prefer Infosys, TCS and TechM among Tier 1 IT; Coforge and Hexaware in mid-tier

The current stock prices of Infosys, TCS and TechM reflect low growth expectations. The stocks trade at ~16X FY2028E earnings and are available at ~4-5% payout yield and ~5-6% FCF yield. TechM has the potential to consistently grow above industry growth. Coforge and Hexaware trade at inexpensive valuations of 18X and 17X FY2028E earnings, respectively. We change the rating of Persistent to REDUCE from SELL, taking into account the reduced downside to our Fair Value.

Quality challengers can have a leg up over incumbents

We believe that incumbents will need to carefully manage revenue deflation due to their large base. Challengers have the luxury of being able to cannibalize themselves in order to get a bigger portion of the pie. Quality mid-tier such as Persistent, Coforge, Hexaware and LTIM will benefit, while Tier 1 IT who are the incumbents will be under pressure. Note that these challengers also have a higher revenue exposure to application services, which is more vulnerable to Gen AI deflation. This needs to be offset by participation in new spend areas and share gains from incumbents.

We adopt a conservative stance in baking higher revenue deflation

The current Gen AI adoption trends tie-in with ~2-2.5% impact, which was already baked in. Higher downside risks can accrue from further improvement in LLM capabilities in software engineering and broader services categories. We take particular note of improvements in agents and agentic AI. Capability improvement and experimentation with agentic software engineering are increasing. Agentic software engineering can drastically accelerate software development. While it has some applications in greenfield software development, we believe that the scope for broader application in implementing commercial software packages, complex system integration and brownfield development is limited as of now. Similarly, application of agentic systems in IMS and non-customer BPO is also limited. The higher deflation figures bake in some risk of broader adoption of AI agents and agentic AI systems.

Latest frontier models show incremental improvement in coding and material jump in other areas

Exhibits 8-10 portray the performance of some of the latest models from the top three frontier AI labs—OpenAI, Anthropic and Google. Models show incremental accuracy improvements in coding and software engineering-related benchmarks and a material jump in a few other areas such as computer use, abstract reasoning and long-horizon professional tasks. Improvements in these other areas can also bring in productivity efficiencies in SDLC and other IT services segments.

It is difficult to gauge how incremental improvements in benchmarks on a high base will shape adoption. For example, it is possible that an increase of accuracy rate from 80% to 90% can catalyze a higher degree of adoption than an increase from, say, 10% to 60%. At the same time, the industry has replaced easier benchmarks with tougher benchmarks due to benchmark saturation.

Difficult to estimate level of productivity possible in IT services given complexities involved

Based on our analysis and channel checks, we see two broad views emerging, which are described below:

The current state-of-the-art coding LLMs and agents may not be sufficient to drive large productivity across real-world software engineering and IT services tasks. However, when combined with fine-tuning, training on enterprise data, ability to use external tools (both AI and deterministic), built-in verification modules and guardrails and smart usage of human-in-the-loop, they can deliver substantial productivity. The productivity levels will only keep improving as Gen AI capabilities and understanding of how to use the models improve.

Enterprise IT is a complex maze of different IT systems and a result of decisions taken over several decades with insufficient documentation. Productivity benefits are possible but to a significantly lower extent than what is advertised. In many cases the risks from hallucinations, lower reliability, lower accuracy and higher costs outweigh productivity benefits.

As of now, we do not have sufficient data to validate either of these two views. We do acknowledge, however, that plenty of capital and talent is being deployed to make the first view a reality. The interest level among developers in experimenting with agents and agentic AI across software development is also high. These factors lead us to assume higher revenue deflation in FY2027-28E and bake in higher risk of disruption despite benign commentary by companies on net Gen AI headwinds in FY2027E.

Expect IT services to continue to be relevant in the long term

While we bake in a higher disruption and lower flow-through of technology spending growth to IT services, we do not agree with the view that IT services companies will not be required or the view that IT services will remain stagnant in the long term. The former aligns with the assumption of zero terminal value, while the latter can be taken aligning with zero terminal period growth. Exhibit 11 indicates the scenario of zero terminal period growth. The target multiple for Tier 1 IT is in the range of 10-13X.

Given the rapid progress in Gen AI capabilities in software engineering and broader IT services and no evidence of any theoretical limit to the extent of productivity possible, these scenarios have a non-zero probability, in our view. There is a large downside to stocks in either of the other scenarios. We believe that the perceived probability of such scenarios occurring may keep fluctuating with new developments in Gen AI and their applications to IT services, leading to volatility in stock prices. Our current belief is that IT services will be relevant in the long term. We do not expect either of these scenarios to play out.

Extent of deflation will depend on the gross impact of productivity savings and reinvestment rate

While Exhibit 1 lays out our assumptions leading to a deflation headwind of about 3%, we add that it is based on several assumptions. First and foremost, we expect the headwinds to be different for various service-lines, with application development and customer BPO being the most vulnerable and IMS and consulting the least. The productivity expectations will evolve with magnitude and direction of improvement of Gen AI capabilities.

Secondly, we expect 50% of the gross impact to be reinvested into IT services in the form of modernization initiatives, higher custom application development, new use cases, etc. While we expect savings to be reinvested in technology, we do not know whether it goes to IT services or whether a good portion gets diverted toward hardware, software and AI labs. The reinvestment rate matters a lot. For example, even if the gross impact were to double, a reinvestment rate of 75% instead of 50% would keep the overall net deflation number constant.

Shift of impact to FY2027-29E compared to FY2026-28E earlier

We estimate the peak impact of deflation to be in FY2028 from FY2027 earlier. We had earlier estimated net impact of 2.6% CAGR spread over three years (FY2026-28). We estimate that the impact in FY2026E has been lower at ~50-100 bps, while capability improvements in software development and broader IT services have continued to improve in the last 9-12 months. Enterprise Gen AI use-case adoption continues to be in early stages. These factors drive our shift of the impact to FY2027-29E.

Measured recovery in demand drives moderated ERD growth over FY2027-28E

R&D spending by clients is likely to be better in CY2026 across most verticals. However, automotive OEMs remain under pressure from weak demand and no material change in customer preferences, further delaying large platform development programs for up to a couple of years. The shift in the nature of demand would benefit companies with relatively higher exposure to body engineering and VAVE capabilities. These programs tend to be much smaller than larger development programs. Further, clients’ priority to optimize R&D spending would imply some pricing pressure. Companies would need to focus on traditional levers to manage profitability. AI adoption would be limited to documentation and validation tasks, and we do not expect material deflation from AI adoption for ERD pure-plays. Aerospace demand continues to be led by after-market services and initiatives by OEMs to improve throughput. We expect the vertical to maintain momentum in the next year. Secular growth drivers from sustainability initiatives in utilities and industrial verticals would continue to drive growth.

Contact services BPO could have higher deflationary impact in the upcoming quarters

Indian pure-play BPO services companies benefited from increased outsourcing and offshoring of processes by clients. Some of this was from in-house G&A spends and more from share gains from competition. However, increased AI adoption by enterprises would increase the asks of the extent of productivity benefits to be passed back, impacting growth. Efficiency measures would enable the companies to manage profitability in the immediate term. On the other hand, while industry-specific BPO would remain relatively less impacted, cost pressures of healthcare clients, specifically those with greater exposure to Medicare Advantage plans would be a drag on medium-term performance for companies such as Sagility. However, we note that improved execution on mining smaller accounts would enable to minimize the drag on overall growth for the company. We raise CoE for pure-play BPO services companies by 50 bps to 12.5% while maintaining terminal growth at 5%. We also moderate our long-term growth assumptions across companies by 100-200 bps over FY2026-36E while also factoring in up to 100 bps moderation in profitability during the period, driving ~19-34% cut to our FVs. We maintain BUY on Sagility and REDUCE on ECLX and FSOL.

Above views are of the author and not of the website kindly read disclaimer