The 2026 IPO Boom Faces Its Liquidity Test

The US IPO market has not seen a first half like this since 2021. Through July 2, American exchanges hosted 194 IPOs raising $155.8 billion, with SpaceX’s $86.2 billion listing alone accounting for more than half that total. Even stripping out the largest IPO in history, the remaining $69.9 billion was more than double the $31.6 billion printed in the same period of 2025{{cite:8efcf52eb15a}}. Dealogic’s IPO Health Index for the Americas now sits near post-2021 highs, and 19 deals have already priced since SpaceX, raising a cumulative $6.9 billion{{cite:8efcf52eb15a}}.

The question occupying every ECM banker is whether this pace survives the next eight weeks. Three distinct liquidity events are converging: SpaceX’s phased insider unlock, a wave of Hong Kong AI IPO lockup expirations, and a second-half pipeline anchored by Anthropic and OpenAI. Each one tests a different facet of demand. Together, they will reveal whether 2026’s issuance boom is built on durable institutional appetite or on scarcity and FOMO.

SK Hynix: The Demand Proof Point

On July 10, SK Hynix completed the largest foreign IPO in US history, raising $26.5 billion by selling 177.9 million American depositary shares at $149 each{{cite:70d0c6fe02d3}}. The deal topped Alibaba’s $25 billion 2014 listing and marked the second-largest US share sale on record, behind only SpaceX{{cite:af52f69f9406}}. The ADRs opened at $170, 14 percent above the offer price, and climbed further in early trading{{cite:af52f69f9406}}.

Demand was reportedly more than seven times the available shares, remarkable for a company that priced at a 2.7 percent premium to its three-day average in Seoul{{cite:70d0c6fe02d3}}. The deal’s success says something specific: investors are willing to pay up for companies whose revenue is directly tied to the AI infrastructure buildout. SK Hynix is one of Nvidia’s primary suppliers of high-bandwidth memory (HBM), the component that makes AI GPUs viable. That supply-chain link mattered more than the so-called Korea Discount that has historically suppressed valuations for chaebol listings{{cite:70d0c6fe02d3}}.

Use of proceeds tells the next part of the story. The capital is earmarked for a new fab in South Korea, a packaging facility, and EUV scanners for next-generation chip production{{cite:70d0c6fe02d3}}. Meanwhile, US Commerce Secretary Howard Lutnick used a Micron event the same week to signal that Washington wants SK Hynix and Samsung to build fabs on American soil, even as both Korean chipmakers pledged more than $550 billion in domestic manufacturing investment{{cite:70d0c6fe02d3}}.

SpaceX’s Phased Unlock: A Ladder, Not a Cliff

SpaceX’s June 11 listing introduced an unusual lockup structure. Instead of the standard 180-day cliff, the company built a tiered, rolling release schedule designed to meter insider selling over time{{cite:7ae40bed5351}}. At launch, less than 5 percent of SpaceX shares were tradable, meaning early price strength reflected artificial scarcity as much as genuine demand{{cite:fab8a0def4e1}}.

The unlock schedule works in stages:

If all tranches clear, insiders could sell as much as 44 percent of SpaceX shares by early September, potentially increasing the tradable float by roughly 900 percent{{cite:fab8a0def4e1}}. The performance hurdle matters: an extra 10 percent becomes available only if the stock trades at least 30 percent above the IPO price. That creates a cleaner test of whether buyers are absorbing supply or running from it{{cite:fab8a0def4e1}}.

The design is deliberate. A phased release spreads selling pressure across weeks rather than concentrating it on a single day. But structure alone does not guarantee a good outcome. Morningstar noted that staggered lockups have not always held up newly listed stocks, citing Facebook’s 2012 lockup period as a cautionary example{{cite:fab8a0def4e1}}. The real test is not how many shares can sell, but how many do sell once the market stops treating the stock like a rarity and starts pricing it like a normal float.

One ECM banker interviewed by Dealogic put the stakes plainly: “If SpaceX starts to trade down that could close the door for other AI issuers”{{cite:8efcf52eb15a}}.

Hong Kong’s $11.5 Billion AI Lockup Wave

While US markets celebrate, Hong Kong is absorbing a different kind of liquidity event. Approximately $11.5 billion in shares from China’s leading AI startups became eligible for sale as lockup periods expired in early July{{cite:a57963ac0116}}. The affected companies include Zhipu AI and MiniMax, two of China’s “Four New Tigers” alongside Moonshot AI and 01.AI{{cite:a57963ac0116}}.

The market response was divergent. Zhipu’s shares actually jumped after its lockup expiry, after nearly 70 percent of cornerstone investors publicly pledged long-term holding{{cite:da0b1838b375}}. MiniMax was not so fortunate: the stock fell 17.98 percent to HK$297.40 on July 9, its first lockup expiry day, wiping out more than HK$20 billion in market value in a single session{{cite:da0b1838b375}}.

The split tells you something useful. When cornerstone investors signal commitment, the market treats the unlock as a non-event. When they do not, the float expansion becomes a selling catalyst. The backdrop compounds the pressure: Chinese AI firms are caught in a domestic LLM price war that ByteDance and Alibaba have aggressively intensified, while US chip export restrictions force companies to train models on less powerful hardware, thinning margins{{cite:a57963ac0116}}.

The Hong Kong experience is a leading indicator of what happens when lockup supply meets fragile sentiment. If the “Four New Tigers” can absorb $11.5 billion in newly tradable shares without a broader sector sell-off, it validates the thesis that AI-related issuance has a durable buyer base. If not, it warns that the same dynamic could play out in US markets when SpaceX’s tranches hit.

The Anthropic/OpenAI Pipeline: Can the Market Hold Two Giants?

Beyond the lockup tests, the second-half pipeline carries its own weight. Both Anthropic and OpenAI have confidentially filed for IPOs, according to Dealogic’s reporting{{cite:8efcf52eb15a}}. OpenAI may delay its listing until 2027, but ECM bankers generally expect Anthropic to come first, potentially in September or October{{cite:8efcf52eb15a}}.

The risk is concentration. One US-based investor warned that having OpenAI and Anthropic list in close succession could strain the market: “Those companies don’t have the idiosyncratic features that the SpaceX setup had”{{cite:8efcf52eb15a}}. SpaceX benefited from retail FOMO, an accelerated Nasdaq-100 inclusion pathway, and a narrative that extended beyond AI into orbital infrastructure. Anthropic and OpenAI are more purely LLM plays, and they arrive at a time when skepticism toward premium LLM providers is growing, even from voices previously counted among the technology’s strongest supporters{{cite:8efcf52eb15a}}.

The base case among ECM practitioners is that there is enough capital to support both listings if underlying markets remain constructive. The bear case is that a SpaceX unlock disappointment in August cascades into weaker sentiment just as Anthropic’s roadshow would begin.

Secondary Market: The Quiet Drain

New issuance is only half the supply equation. The secondary market has been active in its own right. Dollar Tree announced a secondary block trade by Mantle Ridge-affiliated funds alongside a concurrent share buyback in late June{{cite:042931136b11}}. Constellation Energy priced an 11-million-share secondary at $281 per share in early June{{cite:042931136b11}}. Ingram Micro launched a $330 million secondary by its principal stockholder, and Rackspace Technology filed a $250 million at-the-market equity distribution agreement with Goldman Sachs on July 9{{cite:042931136b11}}.

None of these are distress sales. Several are structured with concurrent buybacks, meaning the issuing company is simultaneously absorbing shares. But they add to the total supply the market must digest alongside the IPO pipeline, and they signal that financial sponsors are testing the window for partial exits.

What to Watch Next

Three checkpoints will reveal whether the 2026 IPO window stays open or narrows:

1. SpaceX’s first post-IPO earnings (late July / early August). The first 20 percent unlock tranche triggers after this report. If the stock absorbs the new float without a decisive break below recent support, it validates demand. If each subsequent 7 percent release on August 21 and September 10 brings wider gaps and thinner bids, the phased structure is functioning as an exit lane rather than a ladder.

2. Anthropic’s IPO pricing and aftermarket. As the next mega-deal in the pipeline, Anthropic’s reception will either confirm that investor appetite extends beyond SpaceX’s unique narrative, or reveal that the market is selective about which AI stories it will fund at premium valuations.

3. Hong Kong AI lockup absorption. Zhipu’s resilience and MiniMax’s 18 percent drop already provided a split signal. Watch whether the remaining Tiger lockups (Moonshot, 01.AI) produce broader sector selling or remain company-specific. A systemic AI sell-off in Hong Kong would pressure sentiment for US AI listings.

The base-rate read is that 2026’s IPO market has more durability than the 2021 cycle, because demand is concentrated in companies with actual revenue links to the AI infrastructure buildout rather than speculative growth narratives. But the next eight weeks will test that thesis with real supply, not just scarcity-driven price action. The window is open. Whether it stays open depends on what happens when the float actually arrives.