Markets have oscillated wildly in recent months, the debate raging over how to assess and price SaaS industry vulnerability. A fascinating discourse has erupted across social media, investor briefings and the financial press. Pundits, investors and CEO opinions diverge wildly, from imminent catastrophe to the more complacent “our business model is invulnerable”. In the narrower context of Info Services and Financial Data I’m on the side of “this too shall pass”, rather than the AI myths.
That does not mean wait it out. Far from it, this is a major new technology and AI will profoundly change the industry. The PC and internet did too, but like those technologies, AI will disrupt, forcing innovation (however uncomfortable), but not destroy. Remember, If you believed every panicked commentator the world was due to run out of food (firstly) then oil decades ago.
Bull market and AI myths
Calmer voices have weighed in. Citadel Securities produced a dispassionate, fact based narrative that felt very adult after the breathless nonsense on social media feeds. Credit to Morgan Slade‘s post for pointing it out. Everyone should read the Citadel report in full, but I’d particularly point out the chart on AI Adoption rates. Nothing defeats a hype cycle like an actual look at some data.
Citadel Securities
Following on from the previous piece, I wanted to take a deep dive at some of the beliefs underpinning current market sentiment. If I were to critique most of the myths, the mistake isn’t in assessing what’s happening. In certain areas, AI capabilities are already amazing. The productivity benefits are often already here. But…stating as fact that this trend now accelerates dramatically, even exponentially, needs careful, fuller analysis. The Citadel piece calls out (quote above) some of the limitations to that surge hypothesis.
AI myths. What are the real clangers in this “Financial Data firms are doomed” narrative?
Myth #1. Content isn’t a moat.
You can take LSEG’s 98% figure with a grain of salt (5:30 in the video) but they’re thematically correct. Most financial data firms have a potent mix of proprietary assets. That can be unique content, analytics or models, network effects (e.g. communications, trading or the trade lifecycle), or just deep integration with critical systems and processes. While content is often less proprietary than some argue, it is nonetheless often “non-public”. The effect is the same, you need a commercial agreement to access it. This AI myth “everything is on the internet, just scrape it” narrative is simply ill-informed. You can understand why people with an equities focus make this extrapolation / assumption – see #2.
For many ‘disruptor’ products you even need a software license for the incumbent product to see their content. Read that again. The disruptors’ business models integrate this content, requiring an incumbent’s license! In this respect they’re partners as much as competitors, or “co-opetition”, in the vernacular. News is a great example; no institutional investor/trader doesn’t need Bloomberg, Dow Jones or Reuters News. You need a license for it, it’s not just sitting there waiting to be scraped.
Where I do think the incumbents are complacent is assuming this moat won’t reduce (empty) over time. Executive leadership need to be examining every content set, testing their assumptions on both replaceability and criticality. Some content sets, i.e. estimates, were already vulnerable in other ways.
Incumbents, for example, rarely have robust alternative data strategies and the importance and adoption of this content has surged in recent years. Markets evolve so don’t assume the castle walls are always impregnable (staying with the castle but enough with the moat metaphors). At some point, someone invents gunpowder, or a bailey bridge.
Myth #2. Extrapolating exchange trading into OTC markets (equities into FICC)
Every time I read a doomer headline the examples are the same. They’re always “working on a deal”, IPO, equity research or corporate finance. All those areas have a heavy focus on (public) equities or private companies. Given that most of the writers are or were PE|VC banking types, that’s unsurprising. These areas live and breathe the companies landscape.
I’m pretty sure most writers have no clue what a low % of overall desktop TAM those areas represent. It is simply implausible to argue the exchange trading (equities) and OTC (FICC) worlds are the same. Content needs, market conventions, regulatory, these massive market segments are completely different, but never explicitly mentioned. This “AI is eating the world” needs a few modifiers, narrowing the asset class hypothesis, to be even half-true.
n.b. OTC is an acronym for “Over the counter” referring to much of financial markets that do not trade on exchanges. FICC is a common investment bank acronym for Fixed Income, Currencies and Commodities, referring to their non equities trading businesses.
Let’s think about a real example on content needs. Let’s use today’s markets. You are trying to assess oil supply in real time. Iran is working to interdict both production and distribution. You’re assessing storage levels, shipping routes, which facilities are damaged or just offline, shipping insurance policies, geopolitical posturing. Where there’s the ability to increase production over various time horizons.
Tens of billions of dollars of trading and hedging are repositioning daily. This is one example where one needs real time satellite imagery, geopolitical and energy industry research, shipping news and a streaming markets news feed. No institutional decision is going to be made based on some clever prompt to an AI tool, relying on outdated web info. To truly believe an entire industry is at threat, AI firms have a lot of work to do on content acquisition and integration outside of equities and the WWW.
Myth #3. Vibe Coding for Financial Services.
“I was vibe coding on my couch last night and built a cool collateral management system”. OK, maybe I’m exaggerating but not by much. There’s this amusing narrative that financial services enterprise software can be built while horizontal, while half playing Grand Theft Auto.
It’s great you built a CRM or PMS last night. But it misses the point. Cautious regulators understand that banks are the credit transmission mechanism for the real economy. If you think they’re going to let undocumented AI supervise financial plumbing and infrastructure, you’re in the wrong decade. The bulls seem to think nothing can go wrong. Wait for the first cyber crime incident in the public domain where AI played a malicious role. Increased guardrails and general caution around enterprise adoption feel guaranteed.
Even if agentic AI can deliver all these things safely, there will still be humans in the loop. System cutovers, data migrations, executive sign off. Neither risk nor the cost of change is zero. I think that executive point is worth expanding. Sorbanes-Oxley required executives to sign off on the accuracy of reported financials. Might AI adoption receive the same treatment? Nothing slows down corporate risk taking like personal liability.
Regulators aren’t just going to sit around, and the best way to ensure banks quantify the risks sensibly is through liability. We’ve seen how moral hazard can play a role in risk taking and AI is as great a challenge as new financial instruments. It is certainly one path to consider as regulators grapple with policing a fast changing landscape.
Myth #4 GenAI and probabilistic AI is inherently suitable to financial services and markets
This one perhaps less a myth than a topic that doesn’t get enough mainstream media airtime. For the casual reader, it is important to understand generative AI isn’t an effective calculator of anything. It uses a probabilistic approach. The same inputs do not guarantee the same output. That’s problematic in an industry where data is used to allocate billions.
On the other hand, using deterministic AI, every time you give it the same input, it produces the same output. Generative AI doesn’t do this, and can produce mistakes and hallucinations. There are obvious limitations and risks to this form of AI and, as always, regulators will be most focused on protecting consumers. Hybrid systems will be needed but it feels this area needs much greater scrutiny and oversight. AI tools in production environments require levels of accuracy and consistency not currently seen from Generative AI.
Myth #5 All the risk lies with the incumbents
While I don’t see a near term Blockbuster moment, I certainly see risk. Firms with seat based pricing models will certainly remain priced “ex-growth”. AI will improve productivity, limiting industry headcount expansion.
But what about risks to the disruptors? “Profitable AI firm” is an oxymoron, and market valuations leave little room for error, or patience. There are far more blue sky assumptions (and AI myths) built into AI disruptor valuations than the incumbents. One CEO recently told me he monitored over 80 AI players in his space that had raised >$10M. How many of those firms will exist in the next market cycle? Whose >$800M of savings is that?
I’m not saying that they’ll need Chapter 11, but you can be sure there’ll be consolidation. That’s consolidation between disruptors, incumbents buying disruptors and, maybe, disruptors buying incumbents. At some point the question for a large, Anthropic size disruptor becomes “is it cheaper to license content or buy the owner”
The undoubted winner in all of this is the Financial Services industry, as their vendor base is expanding and user workflows are experiencing unprecedented investment. How the spoils are divided amongst the vendors remains to be seen. What I do know is not all this capex will end well, investments will lead to overcapacity – whether in AI firms, in data centers, or simply in the valuation of the investment vehicles themselves.
Recap
There are many questions we do not know the answer to yet, but clearly the market is making a big bet that the incumbents are successful. Incumbent valuations do not imply bankrutpcy, but they certainly imply rough base cases. The analysis feels asymmetric, with genius innovative AI firms receiving capital while bumbling incumbents are priced ex-growth. Nothing is more guaranteed than the failure of a consensus trade.
Looking further out – there are obviously risks for financial data incumbents. Can/will they reignite their content creation engines? Where does their next phase of growth come from? There are only so many marginal buyers of the same content, but research teams and labs that build new content products are so 2018.
Amongst the incumbents there’s been limted appetite for content creation in the last decade. Sure, everyone jumped late on the Private Markets bandwagon through M&A or partnership, but that’s just moving the same pieces around, not innovating. These firms have forgotten hard fought lessons that difficult-to-replicate content, differentiated content – these are the paths to pricing power and market share growth. Not to mention customer satisfaction.
So while I don’t believe that there’s an imminent “Blockbuster” moment and that share price reactions are overdone, AI will accelerate existing structural challenges. Those teams who understand their core competency is content creation, collection and distribution will succeed. The complacent, believing existing content moats are permanent (and sufficient) will continue to see their growth premia evaporate.
