# Los Angeles Doesn't Have to Build Anything: Odd Lots on Existing Assets, M&A, and the 2% AI Number

Source: Realpha Blog (blog.getrealpha.com)
Original article and charts: https://blog.getrealpha.com/en/blog/oddlots-2026-09-19-a-goldman-m-a-banker-helped-bring-the-olympics-to-/

> Notes from Bloomberg Odd Lots, 2026-09-19, with Gene Sykes — Goldman Sachs co-head of global M&A and president of the US Olympic and Paralympic Committee. How existing venues underwrite a 'privately funded' 2028 LA Games, why this is the biggest M&A year on record, and how to read the claim that only 2% of companies see AI in their EPS. Educational, not investment advice.

Published: 2026-09-19
Locale: en
Tags: Olympics, M&A, AI adoption, media rights, existing assets


![A large existing Los Angeles stadium at dusk, crew installing Olympic wayfinding signs on the original concourse, the view running deep from the upper stands out to the city and freeways](/covers/oddlots-2026-09-19-a-goldman-m-a-banker-helped-bring-the-olympics-to--cover.png)

> Strengthen the root and spend with restraint, and even Heaven cannot make you poor.
>
> —— Xunzi, "Discourse on Heaven" (Warring States period; translation mine)

## What this episode is about

Bloomberg's Odd Lots ran a live recording on 2026-09-19 from the Future Proof conference in Huntington Beach, California. The guest was Gene Sykes: co-head of global M&A at Goldman Sachs, president of the US Olympic and Paralympic Committee, and a member of the International Olympic Committee. Joe Weisenthal opened with a tangent — every time he comes here he wonders how anyone on the westward migration managed to stop in Nebraska when California was a little further on.

One guest wearing both hats splits the conversation in half. The first half is the 2028 Los Angeles Games: who pays, and whether the traffic will swallow everyone. The second half is his view from four decades in dealmaking — this cycle of AI capital spending, the anxiety running through media, and how far enterprises have actually got with AI.

What stayed with me is that both halves run on the same thread: **what you already own**. Los Angeles can claim it doesn't need to build because the stock is already there. Back catalogs get more valuable in an AI world for the same reason. And most companies have been talking about AI for two years while their financials show almost nothing, because wiring a new tool into an old operation is much harder than buying subscriptions.

## The main points

**1. The Olympics and the World Cup are not the same order of magnitude.** Sykes laid the numbers side by side: this year's North American World Cup had 104 matches, 16 host stadiums, roughly 600 athletes, 6.5 million tickets sold. The Los Angeles Olympics and Paralympics together will have 17,000 athletes, around 900 medal events, and 15 million tickets. He described it as maybe 90% of everything you can imagine in the world of sport.

![Two grouped bars compare the events: the World Cup's athlete bar is so short it is nearly invisible and its ticket bar is about a quarter of the Olympic one, while both Olympic bars run full width.](/figures/olympics-vs-world-cup-scale-en.svg)

**2. "First privately funded Games" needs a discount.** Tracy Alloway asked whether the headline she kept reading was true. His answer: no government support to put the Games on, with the exception of the federal government handling security and security-related logistics — with 206 countries sending teams, that part can't happen otherwise. What makes the budget work is that Southern California already has the densest sports infrastructure of any city. Nothing new gets built, and there is no athletes' village: everyone stays at UCLA, which houses more students in a normal year than the Games need to house athletes.

![Two stacked bars compare cost: a typical host city has new venues, an athletes village and games operations, while for Los Angeles the first two become dashed empty outlines and only the bottom block stays solid.](/figures/la-2028-cost-stack-missing-two-blocks-en.svg)

**3. The fix for traffic is telling people in advance.** Joe said queuing and sitting in traffic are two of his least favourite things on earth. Sykes answered with 1984: plenty of people said they would leave town, and the traffic turned out manageable. His mechanism is that people behave according to what they know and expect — if they know where and when, they adjust. Today's high-occupancy vehicle lanes become Olympic lanes.

**4. No corporate advertising inside the field of play.** This rule was new to me. Sykes said the Olympics are known for clean venues: anywhere you see athletes competing, there is no corporate advertising at all, and everyone abides by it. Sponsor visibility lives offstage — the area where an athlete puts a helmet on, for instance. Tracy raised the commercialization complaints aimed at the World Cup, such as water breaks timed around ad inventory, and he said the Olympics won't go there. The money arrives elsewhere: the IOC sells media rights years ahead, NBC holds the United States through 2036 at more than $1.5bn per Games, and nobody yet knows where the 2036 Games will be held.

![Concentric circles: the centre is the field of play with zero advertising, the outer ring holds back-of-house and sponsor visibility, and a leader line runs from outside the circle to broadcast rights.](/figures/clean-venue-money-outside-the-frame-en.svg)

**5. The biggest M&A year on record, with a different engine.** Sykes expects this year to run well past the last peak in 2021. The structure has shifted too: for several years, 40% of the M&A market was private equity sponsors selling companies they owned. This year that's 30%. The reason isn't sellers pulling back — it's substantive demand from industries building out for AI, across natural resources, power generation and semiconductors, with everyone wanting proprietary chip technology.

![Two stacked bars, with the 2026 one clearly taller, where the private equity block shrinks from four tenths to three tenths of the stack and corporate demand carries the rest.](/figures/ma-2026-bigger-pie-different-driver-en.svg)

**6. Enterprise AI adoption: 2%.** Joe floated his theory that enterprise AI adoption is effectively zero. Sykes supplied a number: Goldman's own analysis puts it at 2% today — companies that have said in earnings reports that they see a positive EPS impact from their use of AI. Somewhere between 20% and 30% have actually implemented strategies to use it across what they do. Joe's gloss is the part I liked: every company bought subscriptions, but asking "I'm going on Odd Lots, who are these Joe and Tracy fellows and what do they ask" is glorified search.

**7. Back catalogs get more valuable, and the M in M&A is mostly gone.** On content libraries, Sykes thinks AI will find more uses for existing content, which pushes value up. On dealmaking, Joe relayed a friend's line: it's all A now. Sykes didn't argue — mergers of roughly equal companies come with board seats from both sides and split management teams, and those arrangements exist to get the deal done and let shareholders feel they sold or merged at a premium. A year after closing, decision-making consolidates into one set of hands.

## Going further

### "The news says AI changes everything. Why can't I see it in the financials?"

I feel this gap myself. Every software name I hold spends its earnings call on AI, and then I open the income statement and can't find the line.

Sykes's two numbers pull the gap apart: 2% can point to EPS impact, 20–30% are actually doing the work. A gap of more than ten times says adoption and realisation are two separate things as far as a financial statement is concerned. Plenty of companies use it. Few can turn that use into a number they're willing to state publicly.

![A three-stage funnel: the top stage runs nearly full width for buying subscriptions, the middle narrows to about three tenths for wiring it into the business, and the bottom is a thin sliver for 2%.](/figures/ai-adoption-funnel-two-percent-en.svg)

For reading companies, that means the next time one says it has embraced AI, it's worth asking which box it sits in: bought subscriptions, rewrote workflows, or can point at a cost line that fell or a revenue line that grew. When Sykes described Goldman's own position, he used phrases like getting to conclusions faster and having capacity to think about different kinds of problems — no dollar figure attached. I read that as honesty rather than evasion. He added something sharper: AI doesn't make people smarter, it gives people who are good at synthesizing many independent variables access to more and better information, which gives them more confidence and better judgment. The underlying skill still gates the outcome.

Which makes the distance between 2% and 20% the thing to check against later. If the ratio hasn't moved in a year, whatever is blocking it is harder than it looks. If it moves fast, realisation is under way. That's a rare case of a number you can keep score with.

### "Does this build-out rhyme with the dot-com bubble? Should I be scared?"

Tracy asked this, and asked it well: we laid enormous amounts of cable, the direction was right, the internet did become a thing, it all got used eventually — with a large bubble in between.

Sykes answered in two parts. First, the honest unknown: how big the market ends up, and whether we're committing to too much infrastructure relative to real demand five or six years out — he says we don't know. Second, what he thinks is different this time: the financiers changed. The 2000–2001 build-out was principally venture capital. This one is funded by the most successful large technology platforms in the world, which had hundreds of billions of dollars of free cash flow until this year.

![Two panels side by side: the left pool of money is small and thin, the right pool is far thicker but has a block at the top being eaten away, and both have arrows running down into infrastructure.](/figures/who-pays-vc-versus-free-cash-flow-en.svg)

That paragraph carries good news and bad news at once. Funding from strong free cash flow means the build-out doesn't stop the instant capital markets tighten, so a deflation would take a different shape. But "until this year" finishes the thought: that free cash flow is being consumed by capital spending. Right direction and wrong timing is the expensive way to be wrong — the dot-com investors who saw it correctly and bought in 1999 waited a very long time to get back to even.

![Two lines diverge: one rises smoothly to the right for real demand, the other spikes, collapses and only slowly recovers for market price, with the entry point marked at the spike's peak.](/figures/right-direction-wrong-timing-en.svg)

The method I take from this section is to split "will the demand arrive" from "how fast does this price already assume it arrives". Sykes leans positive on the first and says he doesn't know on the second. Collapsing them into one question turns a long-term view into an entry decision.

### "How much of what a seller says can I trust?"

The structure of this episode is worth noticing. Sykes is selling the Games and works in the deal business. Both claims — LA will run well, this is the biggest M&A year ever — serve him. Joe said so himself at the end: it still sounds like a mess to him, but the guest is a Goldman banker, and if anyone could make a compelling pitch, it's him.

That doesn't mean discounting every sentence. What I do is sort the material into two piles: **checkable facts** and **judgments about the future**. No new venues, athletes at UCLA, NBC through 2036, sponsor share of the deal market going from 40% to 30%, 2% of companies citing EPS impact — those are checkable, and if you can't verify one, you set it aside. "This is going to be the biggest event in the history of the world" and "you'll feel you were part of something remembered forever" belong in the second pile.

The moment a seller carries you along is usually the moment the two piles get mixed: three verifiable numbers and one emotional conclusion in the same breath, and the whole passage sounds solid. I've fallen into this one. Every figure checked out, and the conclusion was still someone else's.

## Worth reading next

- The episode itself: Bloomberg Odd Lots, 2026-09-19, recorded live at the Future Proof conference with Gene Sykes
- For Olympic revenue structure, the IOC's public material on broadcast rights and the TOP sponsorship programme is the starting point; host-city domestic sponsorship is sold separately
- Goldman Sachs publishes an annual M&A outlook, and a public version is available (Tracy read hers on the beach; I'd suggest somewhere with a table)
- For the dot-com comparison, long-run telecom capital spending from 1999–2002 alongside subsequent fibre utilisation data is the useful pairing

## One thing to take with you

Los Angeles can say it doesn't need to build because the venues are already standing, already hosting large events all year, run by people who know exactly what to do. When an opportunity shows up, what decides whether you can catch it is the stock you already have and still have running.

Something I've tried: take a sheet of paper and write down three things you already own and haven't used in six months. A skill, an empty room, a piece of equipment you bought and shelved, a relationship that hasn't broken but has gone quiet. Pick one and use it this week for something you were otherwise going to pay for or spend time on — shoot the photos you were about to outsource with that camera, ask the friend you haven't called the question you're currently searching for. The point is the test: of the stock you think you have, how much of it still runs.