Anatomy of a P3 Megaproject Failure: The $755 Million Swing on I-4 Ultimate

Florida’s largest construction project turned a $255 million projected profit into more than half a billion dollars in losses — and the Eleventh Circuit’s opinion affirming the resulting judgment is the rare court document that shows, step by step, how a megaproject comes apart. A forensic reading.

“Not all ventures strike gold; some strike sinkholes.”

That is the first sentence of the Eleventh Circuit’s April 15, 2026 opinion in Lane Construction Corp. v. Skanska USA Civil Southeast, Inc., No. 24-12638 — and for once it is not judicial color. The joint venture that built Florida’s I-4 Ultimate project literally drilled into a sinkhole so large the team “couldn’t find the bottom,” at the precise location of a bridge that one of its own members had deemed the most critical item on the critical path.

The numbers first. In 2014, Skanska USA Civil Southeast, Granite Construction, and The Lane Construction Corporation formed a joint venture — SGL — to design and build the I-4 Ultimate project: a $2.3 billion reconstruction and widening of a twenty-one-mile stretch of Interstate 4 through Central Florida, spanning seventy-two bridges and fifteen interchanges. The court describes it as the most expensive construction endeavor in state history. SGL’s winning bid projected roughly $255 million in profit. By 2018, three years into construction, the projection had inverted to a $108 million loss. By the time the dispute reached a ten-day bench trial in late 2023, losses exceeded $500 million.

That is a swing of more than $755 million against the original projection — roughly three times the profit the venture set out to earn, in the wrong direction. Lane’s own then-CEO called it “almost unheard of in the United States,” telling the court that “[p]eople don’t lose money like that.”

What makes this case worth careful study is not the loss itself. Megaprojects lose money with depressing regularity, and the cost-growth literature — GAO-20-195G among it — documents the pattern thoroughly. What makes I-4 Ultimate different is that the full anatomy is now public. A forty-nine-page published federal appellate opinion, drawing on a forty-page trial opinion and a ten-day evidentiary record, walks through the project structure, the loss trajectory, the internal deliberations, the legal theories, the settlement mathematics, and the partnership breakdown in granular, quotable detail. For practitioners in project controls, risk, and claims, it is the rarest kind of document: a complete forensic specimen.

The Structure: Where the Risk Actually Lived

The I-4 Ultimate project was procured as a public-private partnership rather than traditional design-bid-build. The Florida Department of Transportation contracted with a private concessionaire — I-4 Mobility Partners OpCo LLC — which agreed to finance the project, secure contractors, and deliver to FDOT’s specifications. In return, the concessionaire would receive milestone payments from the state and a forty-year exclusive maintenance agreement the court values at roughly $75 million per year.

SGL sat one level down, as the design-build contractor under a Design Build Agreement with the concessionaire: complete the project in six years for the $2.3 billion bid price. Late completion meant liquidated damages. Abandonment meant uncapped liability. And Section 20.2 of the DBA stated — captioned in bold, the court notes — “No termination for Extended Relief Event[s].” The court’s summary of this architecture deserves to be read aloud in every bid room pricing a P3 pursuit: while the concessionaire “was legally responsible for ensuring the project’s completion, the brunt of the financial risk was actually borne by SGL.”

The ownership structure added a feature that would later dominate the litigation. The concessionaire was half-owned by Skanska Infrastructure Development — a separate company sharing an ultimate parent, Skanska AB, with SGL’s managing partner. The same corporate family therefore sat on both sides of the design-build contract. This was disclosed: Lane and Granite received the organizational chart more than a year before signing the joint venture agreement, and the court observes that the arrangement “is quite typical among P3 ventures.” Lane conceded it knew the structure from the beginning, and conceded the structure did not itself create a conflict — only “the possibility of conflict.”

That phrase is worth holding onto, because the deal’s economics defined exactly when the possibility would mature. As long as the project performed, contractor and concessionaire interests aligned. Only if things went badly would they diverge. The contract’s risk allocation, in other words, contained a latent fault line whose trigger condition was the one thing no one priced: catastrophic underperformance.

The Data Turned First

Construction began in 2015, and per the court, “almost nothing went according to plan.” A labor shortage and geological hazards drove delay and unforeseen cost. Material prices rose with deteriorating economic conditions. Hurricanes Matthew and Irma, in the court’s narration, “offered no respite.” By 2018 the venture’s projected $255 million profit had become a projected $108 million loss — a $363 million reversal in the performance forecast within roughly three years of a six-year schedule.

One absence in the record is worth flagging, because readers will be tempted to supply it: nowhere in forty-nine pages does the court mention COVID-19. The loss drivers the opinion names — hurricanes, inflation, a labor shortage, geological hazards — all enter the narrative years before 2020, and the forecast had already inverted by 2018. The project ran through 2022, but on this record the trajectory was set, and visible, well before the pandemic began.

Then came the defining event. SGL drilled into a sinkhole at the proposed site of a bridge that Lane itself had deemed “the most critical item on the critical path.” The venture assessed the issue at $48 million and predicted a 245-day delay — and in May 2018 it submitted a relief claim through the contractual mechanism available to it.

That mechanism rewards attention, because it is where project controls and contract administration converge. SGL could not claim against FDOT directly. It had to submit an “Equivalent Claim Notice” to the concessionaire, which had to agree the claim was neither “false or frivolous” nor “based on defective cost and pricing data” before passing it to FDOT as a qualifying Relief Event for the state’s approval. Read that standard again: the contractual gate for relief on a multibillion-dollar program turns explicitly on the integrity of the contractor’s cost and pricing data. Relief, under this structure, is not an argument. It is a documentation discipline — adjudicated by a counterparty whose own incentives are in motion.

The opinion does not publish SGL’s monthly performance data, so the precise shape of the deterioration curve is not in the record. But a $363 million forecast reversal does not arrive in a single reporting period. It accrues — through schedule erosion, productivity variance, escalation, and the compounding interaction of all three — and it surfaces, increment by increment, in exactly the data streams an earned-value system and an integrated master schedule exist to monitor. By 2018 the trajectory was unmistakable enough that the venture’s principals were no longer debating recovery. They were debating escape.

When the Numbers Turn, the Lawyers Get Creative

What happened next is the most instructive — and the most cautionary — stretch of the record.

Recognizing that Relief Event recoveries would “only scratch the surface” of the losses, SGL asked its outside counsel, Holland & Knight, to find a way out of the DBA. The first memorandum was unequivocal: SGL had no termination rights. A second memorandum mapped a theory of leverage — the concessionaire could conditionally terminate its own agreement with FDOT if qualifying relief events delayed the project at least 180 days, and SGL could arguably pressure the concessionaire to invoke that right — but it came wrapped in caveats. A second firm, retained for another opinion, added more: the maneuver would have “serious consequences,” might “push FDOT into self-preservation mode,” could ignite “protracted litigation,” and should be “a remedy of last resort.”

Then, at an October 2018 meeting of the venture’s executive committee in New York, Lane’s outside counsel presented a more aggressive theory built on a single provision extracted from the nearly 200-page DBA. On a literal reading, the clause appeared to say that if the concession agreement terminated and FDOT elected to continue the design-build agreement, the concessionaire would assume SGL’s construction obligations — leaving SGL, in the court’s characterization of the theory, “home free.”

To most of the room this was nonsensical: it would have required a special-purpose financing vehicle to build a megaproject on its own nickel. Skanska USA’s general counsel put it plainly: the provision “didn’t make any sense” and “had to be a typo.” It was. Holland & Knight located the DBA’s drafting attorney, who confirmed the clause contained a scrivener’s error and supplied the corrected text. Properly read, the provision did the opposite of what Lane’s theory required — it implemented FDOT’s right to step in and keep SGL on the job. Holland & Knight’s final analysis called the aggressive positions “unsupportable” and pursuit of the termination request “reckless.” The Eleventh Circuit’s epitaph for the theory runs one line: “no court was going to allow a couple of typos to change that fact.”

The Fork in the Road

By early 2019, termination was — in the record’s words — “on the back burner or even dead,” and the venture pursued the alternative: negotiate with the owner. By July 2019, SGL was nearing a $125 million settlement with FDOT that preserved its right to pursue future relief claims. Lane’s CEO promised in person to honor the venture’s capital calls — then texted that he was “not authorized to agree,” because the decision belonged to the principal of Lane’s Italian parent, WeBuild. In October 2019, Lane and WeBuild formally refused to back the settlement and demanded termination instead. One WeBuild executive’s negotiating philosophy, preserved in the record from the New York meeting: “first you put the knife to the throat and then you negotiate.”

The final agreement with FDOT was executed in April 2020 on a two-to-one vote, Lane dissenting, with Skanska exercising its unilateral authority as managing partner. The terms: $125 million to SGL, a one-year extension of the completion deadline, a dramatic limitation of liquidated-damages exposure, and preservation of future relief claims — in exchange for waiving the termination right. Whatever one thinks of the price, it is a textbook illustration of what negotiated relief on a distressed P3 actually looks like: cash, time, and downside protection, purchased with finality.

Lane paid several capital calls afterward. Then, in January 2021 — after its own employee had circulated the draft $21 million working-capital request, and its own executive-committee representative had written “I review it and it looks ok to me” — Lane refused to fund, sued Skanska for breach of fiduciary duty, and declared in writing that it would make no further capital contributions. A Lane representative later put it plainly in deposition: “At this point, Lane was not going to pay this call. We were heading toward litigation.” By August 2021, Lane’s unpaid share approached $60 million. Skanska and Granite covered the difference to keep the project funded. The highway was completed in 2022. The lawyers carried on for four more years.

What the Courts Decided — and What They Never Had To

The district court resolved the contract question on summary judgment with a sentence every joint-venture partner should memorize: the “contract and fiduciary claims are different animals.” A partner who believes its co-venturer has breached a fiduciary duty does not thereby acquire the right to stop funding the venture. Self-help by withheld capital is itself the breach. The agreement then enforced itself: upon default, the JVA stripped Lane of its seat on the venture’s Executive Committee — which made every subsequent capital call, approved by the two remaining members, “unanimous” under the contract’s own terms.

After the ten-day bench trial, the district court went further, fully absolving Skanska: rejecting the termination request was “a no-brainer, not a breach.” Its catalogue of what even threatening termination would have cost is its own education in megaproject risk: near-certain default exposure, years of litigation with both the concessionaire and FDOT, disclosure of the default as “a permanent black mark” on every future pursuit, and “persona non grata” status with a customer of FDOT’s size. The court ordered Lane to pay $49 million to Skanska and $30 million to Granite — the amounts required to rebalance the partners’ proportionate funding obligations — plus roughly $13.5 million in prejudgment interest. On April 15, 2026, the Eleventh Circuit affirmed in full.

The appellate opinion repays close reading on three points. First, the panel held that the joint venture agreement never modified the statutory duty of loyalty under Florida’s partnership act — and paused to register genuine surprise: “Given the sheer magnitude of the joint venture and Skanska’s looming potential conflict known from the very beginning, it is surprising the JVA failed to even conceive of a conflict, let alone institute safeguards.” Florida law expressly permits partners to engineer conflict carve-outs and cleansing procedures up front. SGL’s 2014 agreement, governing a $2.3 billion venture with a disclosed cross-ownership structure, contained none.

Second, the panel held there was no cognizable adversity. A partner with a different economic stake is not a partner with an adverse interest unless the competing stake pulls toward a departing course of action — and here, every adviser, every co-venturer, the contract’s drafting attorney, the trial court, and the appellate court agreed that rejecting termination was the only sensible course. Granite’s position carried particular weight: a partner “essentially in the same shoes as Lane,” with no affiliate on the concessionaire side, voted against termination, funded the shortfall, and joined Skanska’s side of the litigation.

Third — and of real consequence for the P3 market — the court flagged, without resolving, whether Florida partnership law admits a “fairness” defense to conflicted-interest claims, noting that the model act added one in 2013 and Florida never has. That open question now waits for the Florida Supreme Court, and every P3 design-build venture organized under Florida law is living with the ambiguity in the meantime. (Judge Newsom concurred separately, declining to join that discussion as unnecessary to the result — a reminder that the fairness-defense commentary is dicta, not holding.)

Notice, finally, what five years of federal litigation never adjudicated: a single schedule delay, a single cost variance, a single dollar of the underlying $500 million loss. The courts allocated the loss among partners. The loss itself — its causes, its detectability, its preventability — was settled long before anyone reached a courtroom, in the places such things are always settled: the performance data, the schedule network, and the contract’s risk allocation, between 2014 and 2019.

The Forensic Reading

Strip the litigation away and the record yields five lessons in controls discipline.

Implications for the P3 Pipeline

The Eleventh Circuit went out of its way to note that overlapping contractor-concessionaire roles are common to the P3 delivery model, and it declined to treat them as inherently disloyal. The model survives this case intact — which means the burden it implies falls entirely on participants. The United States has a deep pipeline of multibillion-dollar P3 and mega-program delivery ahead of it, across highway, transit, rail, and — increasingly — federal energy and AI-infrastructure construction. None of those programs is predestined to follow I-4’s trajectory. But all of them share its structural features: concentrated contractor risk, relief mechanisms gated on data integrity, joint-venture governance under stress, and performance data that will tell the truth earlier than anyone wants to hear it.

The difference between a program that absorbs a sinkhole and a program that becomes a published cautionary opinion is not luck. It is whether someone is reading the data with forensic rigor while the window to act is still open.