The week of July 21, 2026, produced two financial market stories that superficially appear unrelated — a massive bond sale for an AI data centre in El Paso, Texas, and record first-half profits from a Swiss wealth manager in Zurich — but that together illuminate the defining dynamics shaping global capital flows in 2026. In the first story, the world’s largest asset manager raised $12.3 billion in debt to fund infrastructure for the world’s most valuable social media company, and bond investors asked for more money than before despite — or rather because of — the frenzied AI capital spending cycle driving demand. In the second story, the same AI-driven market volatility and heightened client activity that is making tech companies spend trillions also produced a record performance for a private bank whose wealthy clients were actively repositioning their portfolios through one of the most turbulent years in recent financial history. The connecting tissue is the sheer volume and velocity of capital now moving through global financial markets in service of a single technological bet: that artificial intelligence will be transformative enough to justify the largest sustained infrastructure investment in human commercial history.
Two Stories, One Underlying Theme
The week of July 21, 2026, produced two financial market stories that superficially appear unrelated — a massive bond sale for an AI data centre in El Paso, Texas, and record first-half profits from a Swiss wealth manager in Zurich — but that together illuminate the defining dynamics shaping global capital flows in 2026. In the first story, the world’s largest asset manager raised $12.3 billion in debt to fund infrastructure for the world’s most valuable social media company, and bond investors asked for more money than before despite — or rather because of — the frenzied AI capital spending cycle driving demand. In the second story, the same AI-driven market volatility and heightened client activity that is making tech companies spend trillions also produced a record performance for a private bank whose wealthy clients were actively repositioning their portfolios through one of the most turbulent years in recent financial history. The connecting tissue is the sheer volume and velocity of capital now moving through global financial markets in service of a single technological bet: that artificial intelligence will be transformative enough to justify the largest sustained infrastructure investment in human commercial history.
Part One: Meta, BlackRock, and the $12.3 Billion El Paso Bond
On Monday, July 21, 2026, Bloomberg reported that BlackRock Inc. was seeking to sell more than $12 billion in bonds to finance a Meta Platforms data centre campus in El Paso, Texas. Four days later, on Friday, July 24, a second Bloomberg report confirmed that BlackRock had formally kicked off the marketing of $12.3 billion in high-grade bonds through a holding company called Sopaipilla Investor, with notes due in 2048 — 22-year paper backing infrastructure intended to be operational for at least that duration. JPMorgan Chase and Morgan Stanley were mandated to arrange the sale and run investor calls. The pricing was expected as early as the start of the week of July 27.
The El Paso project is a roughly one-gigawatt data centre campus that Meta expects to come online in 2028, supporting more than 300 on-site jobs. The holding company issuing the debt — Sopaipilla Investor, tied to Project Sopaipilla Holdings — is owned 80 percent by BlackRock through two of its subsidiaries, Global Infrastructure Partners and HPS Investment Partners, both of which BlackRock has acquired in recent years as part of its aggressive build-out of alternative asset management capabilities. Meta owns the remaining 20 percent. Meta will lease the campus back from the BlackRock-led entity under a long-term arrangement, meaning the campus debt sits on the joint venture’s balance sheet rather than appearing directly on Meta’s own balance sheet — a structure explicitly designed to keep Meta’s reported financial leverage lower than the true economic scale of its infrastructure commitment.
The Structural Template: How Hyperion Created the Playbook
To understand the El Paso deal, it is essential to understand the deal it was designed to replicate. Meta’s Hyperion data centre, located in Richland Parish, Louisiana, was financed in October 2025 through an identical structural template: a joint venture 80 percent owned by Blue Owl, a private-credit firm, and 20 percent owned by Meta, which issued $27 billion in bonds — the largest private-debt offering in recorded history. BlackRock was itself a major investor in Hyperion, buying more than $3 billion of those bonds. The Hyperion template proved sufficiently successful — for Meta, for the infrastructure investor community, and for the bond underwriters — that it was immediately adopted as the model for what Sopaipilla represents: the second iteration of a financing architecture that allows tech companies to build data centre capacity at a scale that could not be placed on any single corporate balance sheet without alarming equity investors and credit rating agencies.
Hyperion has since grown dramatically. In July 2026, Meta announced it was expanding Hyperion to five gigawatts of compute capacity, pushing the estimated cost of that single project above $50 billion — nearly double the $27 billion at which it was originally priced. Combined with the Sopaipilla financing, Meta’s off-balance-sheet data centre commitments through BlackRock-backed structures alone are approaching $75 billion. A Nikkei analysis put the off-balance-sheet debt of the five largest US technology companies — Alphabet, Microsoft, Amazon, Meta, and Oracle — at approximately $1.65 trillion, meaningfully above the $1.35 trillion they report directly on their own balance sheets. The true infrastructure footprint of Big Tech’s AI ambitions is therefore substantially larger than the figures disclosed in quarterly earnings filings suggest, a gap that credit analysts and institutional investors have begun paying closer attention to as the volume of AI-linked debt issuance reaches historically unprecedented levels.
The Rising Price Tag: Why Bond Investors Want More
The central news from Friday’s Financial Times report — the development that caused Meta’s stock to fall modestly in pre-market trading — was not simply the existence of the $12.3 billion bond sale. It was the price. Early discussions indicated bond investors were seeking yields above 7 percent, roughly 0.4 percentage points higher than the yield Meta’s Hyperion structure had secured in October 2025. Pricing was still subject to change before formal launch, but the directional signal was clear: the capital market for AI infrastructure debt had tightened.
A 0.4 percentage point increase may appear small in isolation. At the scale of $12.3 billion in bonds, it translates into tens of millions of dollars in additional annual interest expense. More significantly, it represents a measurable shift in investor sentiment toward AI infrastructure debt as an asset class — and the direction of that shift, even at this early stage, deserves careful attention.
Several distinct factors are converging to produce it. The first is market saturation and investor fatigue. More than $20 billion of high-yield bonds and loans tied to data centre and AI infrastructure financing had been launched in the three weeks immediately preceding the Sopaipilla offering. The volume was becoming difficult to absorb at previous pricing levels. Morgan Stanley expected global AI-related debt issuance to reach approximately $570 billion in 2026 as a full-year figure, with approximately $236 billion already priced by the end of May — roughly four times the pace seen a year earlier. Combined, Alphabet, Amazon, Meta, Microsoft, and Oracle had issued approximately $121 billion of bonds during 2025 alone, compared with roughly $40 billion in 2020. The market was being asked to absorb, in a single year, three times as much tech-company debt as it was digesting five years ago, with 2026 set to dwarf even that.
The second factor is the secondary market performance of earlier AI infrastructure bonds. Bonds from the original Hyperion deal were trading at approximately 96 cents on the dollar as of Thursday, July 23 — meaning investors who bought Hyperion bonds at par in October 2025 had sustained a paper mark-to-market loss of approximately four points in nine months. That performance sends a clear pricing signal to investors considering the Sopaipilla offering: the secondary market for AI data centre bonds has softened since the asset class was new and attracting a yield premium simply for being novel. New issuance needs to offer a genuine incremental yield to compete.
The third factor is growing broader market concern about whether the returns from AI infrastructure spending will justify its costs at the scale currently being deployed. JPMorgan strategists estimated in June 2026 that hyperscalers will spend roughly $5.5 trillion on AI through 2030, much of it borrowed. The question of whether that spending generates sufficient revenue — through cloud services, AI model licensing, advertising efficiency gains, and new product categories — to service $5.5 trillion in accumulated debt and provide adequate equity returns is one that the investment community has not yet answered with confidence. That uncertainty is precisely what gets priced into bond yields as a risk premium.
BlackRock’s Strategic Position: Fee Engine, Not Balance Sheet Risk
For BlackRock specifically, the Sopaipilla transaction has a different meaning from the headline bond sale figures. The $12.3 billion in bonds are not being issued off BlackRock’s own corporate balance sheet — they are being issued by a holding company that owns BlackRock’s infrastructure investment position. BlackRock earns management fees on the capital it manages through Global Infrastructure Partners and HPS Investment Partners; the default risk on the bonds sits with the bond investors who buy them, not with BlackRock itself. BlackRock’s equity stake in the project — worth approximately $2.4 billion at the 80 percent ownership fraction of a $3 billion equity component — creates performance-related upside if the project succeeds, but the enormous debt pile funding the campus is a bond investor’s problem in the event of any adverse outcome.
This is the business model that BlackRock’s CEO Larry Fink has been explicitly building toward through the firm’s aggressive acquisition of alternative asset managers: GIP in a $12.5 billion deal, HPS Investment Partners in a $12 billion deal, and Aligned Data Centers in a $40 billion transaction. Together these acquisitions position BlackRock as a central intermediary in the channelling of institutional capital — pension funds, sovereign wealth funds, insurance companies — toward the infrastructure investments that the AI buildout requires at a scale that no government or single corporation can fund alone. The management fees on $1 to $2 trillion of assets under management in this infrastructure category will be, for the better part of a decade, among the most structurally durable revenue streams in global asset management.
Part Two: Julius Baer’s Record Half-Year — and the Cloud Hanging Over It
On the same morning that the BlackRock-Meta bond marketing was getting underway, a different and equally instructive financial story was playing out in Zurich. Julius Baer Group, the Swiss wealth management bank, announced first-half 2026 results on Tuesday, July 21 that at first glance look like a straightforward success story: IFRS net profit of CHF 673 million ($831 million) for the six months to June 30, 2026, up 128 percent from CHF 295 million in the same period a year earlier — the highest net profit in the bank’s history for any six-month reporting period.
The CEO framing matched the numbers. Stefan Bollinger, who took over as Chief Executive from Philipp Rickenbacher in late 2024 following the Signa-related crisis, described the results as “a solid start to our new three-year strategic cycle,” attributing the record to “pronounced client activity, all-time high assets under management, sustained net new money inflows, and continued improvement in operating leverage.”
The numbers behind the headline support that framing. IFRS operating income reached CHF 2.276 billion, up 26 percent from CHF 1.810 billion a year earlier, driven by higher net commission and fee income, increased income from financial instruments at fair value, and an 80 percent year-on-year rise in net interest income to CHF 130 million. Operating expenses grew by only 2 percent, generating the operating leverage Bollinger highlighted: a business where revenue grows 26 percent but costs grow only 2 percent. The adjusted cost-to-income ratio improved to 62.6 percent from an underlying 62.8 percent a year earlier. Assets under management rose 5 percent year to date to an all-time high of CHF 547 billion, supported by positive market performance, currency effects, and net new money inflows of CHF 5.7 billion. The Common Equity Tier 1 capital ratio stood at 18.5 percent at June 30, up from 17.4 percent a year earlier, and substantially above regulatory minimums.
The Signa Shadow and the FINMA Silence
The 128 percent net profit increase, while genuinely impressive on an absolute basis, requires careful contextualisation. The comparable first-half 2025 period was significantly depressed by two one-off negative items: a mark-to-market net impact of CHF 99 million from the sale of Julius Baer’s Brazilian operations, and elevated net credit losses from Signa-related exposure. The Signa Group was a European property conglomerate built by Austrian businessman René Benko whose 2023 collapse became one of the largest European insolvency events in decades, and Julius Baer had extended substantial loans to Signa entities that turned into credit losses of CHF 606 million — a crisis that cost the bank its previous CEO, attracted the attention of FINMA, the Swiss Financial Market Supervisory Authority, and required a comprehensive overhaul of the bank’s risk and compliance framework.
On an adjusted basis that strips out the 2025 extraordinary items to compare like-for-like operating performance, net profit grew 32 percent from CHF 511 million in the underlying first half of 2025 to CHF 673 million — still a strong and meaningful improvement, but considerably less dramatic than the headline 128 percent suggests to an uninformed reader.
More significant for the market’s reaction than the headline profit number was what Julius Baer did not say. Investors had been anticipating an update on the status of FINMA’s regulatory investigation into the bank’s risk governance failures that led to the Signa losses. No such update was provided in the earnings release. The absence of that update — which would have clarified whether FINMA’s investigation had been concluded, and if so on what terms — was the primary driver of a 4.1 to 5.1 percent fall in the bank’s shares in Zurich trading on results day, as investors priced in the regulatory uncertainty that the silence maintained.
Bollinger and the bank have been consistent on their remediation approach: the revised risk and compliance framework is expected to continue weighing on net new money inflows throughout 2026 and into 2027, before normalising in 2028. That is a meaningful admission — it means the bank is sacrificing some organic growth in the near term, specifically in terms of client leverage and asset-backed lending, in order to rebuild the credibility of its risk management processes in the eyes of regulators and clients alike. The reduction in the number of relationship managers noted in the results — a figure that raised questions among analysts about the bank’s capacity to grow its client base even once the de-risking headwinds subside — adds a further dimension of uncertainty to the medium-term growth picture.
Why Julius Baer’s Performance Matters Beyond Switzerland
Julius Baer is not a globally systemic institution in the way that UBS or Deutsche Bank are. Its CHF 547 billion in assets under management is substantial but not enormous by the standards of the largest global wealth managers. Why, then, does its performance matter as a signal about the broader financial environment?
The answer lies in what Julius Baer’s client base represents. The bank serves ultra-high-net-worth individuals — typically those with at least CHF 1 million in investable assets, and in practice often multiples of that — across Europe, Latin America, Asia, and the Middle East. When this demographic is “releveraging” — increasing the use of borrowed capital within their investment portfolios, a behaviour that Julius Baer reported resumed in May and June 2026 after a pause in the first four months of the year — it indicates that wealthy investors are moving from risk-off, defensive positioning toward a more confident, return-seeking posture. That transition in sentiment among the world’s wealthiest investors both reflects and reinforces the broader market direction: it provides demand for the high-grade bonds that BlackRock is selling to finance Meta’s data centres, it drives trading volumes in equities that generate commission income for banks like Julius Baer, and it channels capital toward exactly the AI infrastructure and related equities that the technology sector’s extraordinary spending plans depend on for continued valuations support.
The resumption of client releveraging in May and June is therefore not merely a detail in Julius Baer’s half-year results. It is an indicator that the global wealth management community — the population of people who have the most capital and the most immediate access to sophisticated investment advice — spent the first four months of 2026 pulling back from leverage amid the Iran war’s market disruption and then, as the ceasefire took shape and market conditions stabilised, moved back toward the leveraged, return-oriented investment postures that characterise bull market conditions. The timing of that shift, and what it means for the rest of 2026, is among the most watched signals in global finance.
The Big Picture: $5.5 Trillion, 0.4 Percentage Points, and What Comes Next
Taken together, the Meta-BlackRock bond deal and the Julius Baer results describe a global financial system that is currently in the early stages of digesting the largest infrastructure investment cycle in history. JPMorgan’s $5.5 trillion AI capex estimate through 2030 implies approximately $1.1 trillion per year of capital deployment — a figure that is simultaneously enormous and subject to significant uncertainty about whether the revenue streams being built will justify the cost. The 0.4 percentage point increase in the yield that bond investors are seeking for Sopaipilla relative to Hyperion is, in this context, not simply a data point about one bond deal. It is an early signal of the market’s own uncertainty about that question, expressed in the only language that capital markets actually speak: the price of money.
For Meta, the higher borrowing cost is a manageable inconvenience at the current scale — tens of millions of dollars of additional annual interest against a company generating over $60 billion in annual operating income. For the broader AI infrastructure financing ecosystem, however, a sustained upward drift in the yield demanded for data centre debt — driven by investor fatigue, secondary market softness, and growing uncertainty about AI return on investment timelines — would meaningfully increase the cost of the entire buildout, and potentially slow the pace at which the next generation of AI capacity comes online.
Whether that repricing proves to be a healthy normalisation of an asset class that had initially attracted too-thin yields for its actual risk profile, or the beginning of a more significant market reassessment of the AI investment thesis, is the financial question of the second half of 2026. Julius Baer’s wealthy clients, who resumed leveraging in May and June, are placing their bets on the former. The bond investors seeking 7 percent yields on Sopaipilla paper are pricing in at least some probability of the latter.
Sources
- Bloomberg — BlackRock Plans Over $12 Billion Bond Sale for Texas Data Center Project
- Bloomberg — BlackRock Kicks Off $12.3 Billion Bond Sale for Meta Data Center
- Bloomberg — Julius Baer Profit Doubles as Clients Add $7 Billion New Money
- Yahoo Finance / FT — Meta’s AI Borrowing Costs Rise on $12 Billion Data Center Deal
- Yahoo Finance — BlackRock Raising $12 Billion in Bonds for Meta Texas Data Center
- Yahoo Finance — Julius Baer First Half Profit More Than Doubles on Strong Inflows
- Quartz — BlackRock Raising $12 Billion in Bonds for Meta Texas Data Center
- Unite.ai — BlackRock Bankrolls Meta’s El Paso Data Center in $12B Debt Sale
- TheStreet — BlackRock Raises $12 Billion for Meta AI Data Center in Texas
- Stocktwits — Meta’s Latest AI Data Center Debt Raise Could Reportedly Cost More Than Its 2025 Record Deal
- Foreign Policy Journal — Meta Platforms Faces Higher Borrowing Costs on $12 Billion BlackRock-Backed AI Data Center Deal
- Julius Baer Group — Press Release: H1 2026 Results — Record Net Profit of CHF 673 Million
- WealthBriefing — Julius Baer Net Profit Doubles to New Record in H1 2026
- Finews — Julius Baer Posts Record Profit — But One Key Challenge Remains
- Investing.com — Julius Baer H1 Profit More Than Doubles, But Shares Dip on De-Risking Warning
- Private Banker International — Julius Baer First Half Profit More Than Doubles on Strong Inflows