Hype Out, Infrastructure In: Crypto's Partnership Rotation

Introduction

We ingested and tagged 5,203 crypto partnership announcements between January 2021 and July 2026, from more than 5,000 news sources: crypto media, mainstream outlets, newswire press releases, project blogs, and company statements. The 3,749 announcements that pair a token with a company are the focus of this report.

Over five years, the composition of that deal flow has almost entirely turned over. For anyone deciding who to partner with next, the useful question is not who the biggest names are but which categories are growing, which specific counterparties are still active, and what a deal actually does once announced.

Two terms are used precisely throughout. A traditional partner is any company from outside crypto: banks, payment networks, big tech, governments, universities, corporate treasuries, and other non-crypto corporates. TradFi is the finance subset of that group: banks, asset managers, brokerages, and payment networks. So every TradFi deal is traditional, but a Microsoft or a government entity is traditional without being TradFi. "Traditional share" counts announcements with at least one traditional partner.

Hype out, infrastructure in

Ranking all 21 partner categories by deal count each year, the 2021 leaders were celebrities, sports franchises, and NFT projects. By 2026, they have been displaced (Table 1):

Figure 1 tracks every category across the six years. Entertainment and sports partners fell from 14% of all deals in 2021 to 5% in 2026. Total partnership activity did not collapse, but the mix changed. The two categories that rose furthest each deserve a closer look, starting with the stronger of the two.

Tokenization went from nothing to over a third of what banks sign

Figure 2 shows the share climbing year by year. TradFi partners signed 85 crypto partnerships across 2021 and 2022, and not one involved tokenization. This year, tokenization is 36% of everything that group signs. On the crypto side, dedicated RWA platforms went from 3 deals in 2021 to 52 in 2024 and 46 in 2025.

Crypto's AI wave started late, jumped more than tenfold, then held

Figure 3 splits the wave by who signed it. Crypto signed 5 AI-themed partnerships in all of 2023. Then came 57 in 2024, 63 in 2025, and 38 through July 2026. Given that ChatGPT launched in November 2022, crypto's AI partnership wave began roughly a year and a half later. The trend has been mainly led by crypto-native projects (e.g., ChainGPT, Fetch.ai, Allora), with non-crypto firms (e.g., Microsoft, Alphabet, IBM) making up 30%.

Institutional involvement grew through every market regime, despite a constant share of the pie

Partnership activity nearly tripled, from 369 announcements in 2021 to 999 in 2025. More telling, 2025 also set the record for companies from outside crypto doing their first-ever crypto partnership: 214 first-time entrants, up from 142 in 2024. Among TradFi names specifically, first-timers went from 38 in 2021 to 67 in 2025.

But the traditional share of deal flow did not rise. Cut by market regime, it barely moves at all (Table 2):

Figure 4 shows both at once. Deal volume rose through all four periods without a single down-step, including the collapse, and the post-election rate is 2.1 times the 2021 bull market. The traditional share sat between 24.7% and 29.9% the whole way, and each of those readings carries a margin of roughly 2 points either direction, so the periods are not meaningfully different from one another. The interesting part is what never happened: the share held up through Terra-Luna and FTX, when crypto-native activity was the worst hit, and it finished no higher after the election than it had been in the 2021 bull market. In absolute terms, traditional partners never paused: 129 deals per year in the 2021 bull, 153 per year through the collapse, and 261 per year now. They did not take the share because everything grew together. But they did not retreat either, and, given what those 18 months did to the rest of the market, not retreating is the more surprising half.

TradFi nearly tripled and stablecoin deals grew roughly 40-fold after the election

Total activity is the least interesting part of the post-election picture, since it was already rising. What changed is which categories grew. Comparing deals per year in the 2021 bull market against the post-election period (Table 3, charted in Figure 5):

The growth is concentrated in the categories that depend on regulatory posture (e.g., TradFi, stablecoins, corporate treasuries, compliance, government), while the consumer and NFT categories that led 2021 continued to shrink. Stablecoin partnerships went from under 1 per year to 27 per year.

Two caveats on reading this as a "Trump effect." The deal rate was already climbing through the spot-ETF period before the election, so some of this is cycle rather than policy. And the post-election window is only 21 months against 16 months for the bull comparison: long enough to be indicative, not long enough to separate a regime change from a continuing recovery. The compositional shift toward regulated-adjacent counterparties is clear either way, whether the driver is policy or the tail of the recovery that preceded it.

Who is actually accelerating

Exchanges still lead raw counts, but three of the 14 busiest partners are not crypto-native. A growing trend of institutional interest is more apparent when we take a look at each partner's deal count by year (Table 4, plotted in Figure 6):

Two things follow, and they cut against the easy reading.

Payments and big tech are not new arrivals. Mastercard, Alphabet, and Visa have each been partnering with crypto since 2021 or 2022, through the last cycle, the collapse, and the recovery. Whatever is happening now, it is not their entrance.

Nor are they uniformly accelerating. Among the traditional names, only Mastercard and BlackRock are clearly increasing. Alphabet peaked in 2024 at 11 deals and has fallen since; Visa has run flat at two to five per year for five years. At the category level, Big Tech signed about half as many partnerships in the last 12 months as in the 12 before. "Institutions are coming" is true of asset managers and of Mastercard; it is not true of big tech, which was already here and is now doing less.

The genuinely new tier is asset management: BlackRock 92% of its deals since 2024, Franklin Templeton 100%, and Cantor Fitzgerald 100%. That is a two-year-old phenomenon, not a five-year trend.

Beneath those names, the market is thin: 2,792 distinct partners appear in the data, and 81% of them appear exactly once. Repeat partnering is rare and concentrated in exchanges.

What an announcement actually moves

Partnership announcements generally do not move price or trading volume. What they move is attention, and the magnitude rises steadily as the market cap falls.

Since every announcement is timestamped and entity-tagged, we also measured what follows one: 4,292 events across 569 tokens, with daily price, volume, and tweet volume around each.

One thing moves, and only on smaller tokens: attention. On the day a partnership is announced, we compared each token's tweet volume against its own normal level for that day of the week, across the 2,592 events with a clean market-cap read once stablecoins, wrapped assets, BTC, and ETH are set aside (Table 5, charted in Figure 7):

Below $2B, a partnership announcement lifts a token's chatter by about a tenth for the day. Above it, nothing moves: the largest tokens are talked about at the same rate whether they announce a partnership or not. The pattern holds in order across all four size bands, and the two halves differ by about 10 percentage points, though with this many announcements we would call that indicative rather than settled.

Methodology. For each announcement, we take the token's tweet count on the day and divide it by that same token's median tweet count for the same day of the week over the prior 30 days, so a Tuesday announcement is judged against that token's ordinary Tuesdays. A reading of 1.0 means a normal day; the table reports the median of these readings across the 2,592 announcements in each band, winsorized at the 1st/99th percentile. The day-of-week matching is the load-bearing choice: crypto is quieter at weekends and 93% of announcements land midweek, so measuring against "the last few days" flatters every announcement. Our first pass did exactly that and produced a +5% lift that turned out to be the calendar, not the news. On this construction an ordinary day scores exactly 1.0, its own normal level, which is how we know the yardstick is fair.

Price and volume are the other half of the answer, and there the result is nothing measurable. Both are indistinguishable from a normal day, and that holds across partner type, market-cap band, individual marquee partner, market regime and partner category. It differs from exchange listings, where trading volume stays elevated for weeks even as the price gives the initial pop back, and the reason is mechanical: a listing changes who is able to buy a token, a partnership does not.

One caveat is worth stating plainly. Measured against Bitcoin, these tokens sit 0.91% lower a week after an announcement, which reads like a decline. But the same tokens run about 0.7% below Bitcoin in any ordinary week too, announcement or not. Nearly all of that gap is their usual underperformance, not the news.

The attention pattern is visible in the individual deals (Table 6). We picked these 10 rows for the partner's name, not for how the token did afterwards, so read the price columns as context, not as a scoreboard:

Every one of these is a recognizable institution, and the tokens run $0.1B to $3.7B: four in the small band under $1B, three in each of the two mid bands, and none in the large band above $10B. But attention does not order itself by size within a hand-picked set: Ondo at $2.9B drew more than Audius at $0.1B. The band gradient above, not this table, is the evidence for the size effect.

The two price columns are the more useful contrast. The peak column can only ever look good, because a week's best price is never a loss; the close column shows how little of that survives. LayerZero peaked 26.6% above BTC and finished 20.3% below it. Seven of the 10 closed negative, and six gave up more than 10 points from their peak. Reading the peak column as a result is exactly the mistake this table exists to prevent: these are illustrations of what the data contains, not a ranking, and a ranking built on that column would not hold up on the next batch of deals.

It is worth following the quietest row in that table past its announcement. SWIFT and Chainlink drew the smallest attention spike of the 10, +132%, and closed 2.5% below BTC. It is also the deal with the clearest institutional follow-through: in November 2024, SWIFT, UBS Asset Management, and Chainlink completed a pilot settling tokenized fund subscriptions and redemptions over SWIFT messaging, and the work carried into 2025 as a transfer-agent standard with UBS as the first asset manager to adopt it. Swift's head of strategy, Jonathan Ehrenfeld, put the requirement plainly: digital assets "must seamlessly integrate with both existing payment systems and digital currencies."

The loudest row had substance behind it too. Two months after Deutsche Telekom took its CELO position, a16z delegated CELO to the validator group Deutsche Telekom's T-Systems MMS was running. Announcement-day attention separated these two deals by a factor of 25, and in neither case did it indicate how far the working relationship would go.

Conclusion

So the value in partnership data is not as a trading signal. For a team deciding who to partner with, it is the composition read above: which categories of counterparty are actually growing, which specific names are open to a deal right now versus which have gone quiet, and how rarely a first deal becomes a second one. A one-off announcement is far easier to win than a lasting relationship. A few deals do coincide with large price moves, as the examples above show, but across thousands of announcements the typical deal does not move the token at all.

Methodology notes

The dataset is partnership-tagged news with a token subject and a company counterparty. The landscape covers 2021-01-01 to 2026-07-31; the market-outcome study stops at 2026-06-30 because a +30-day event window needs 30 days of post-announcement data to exist. Volume and attention are indexed to each event's own 30-day pre-announcement median for the same day type; returns are BTC-adjusted; medians are winsorized at 1/99 and reported with standard errors; delisted tokens are retained. Full event-study detail, per-cut results and limitations are available on request.