The website is open. For the rest of the summer, you won’t need a password to access epistrophy.beehiiv.com. It’s an experiment to make it easier for you to see these notes and what has come before this.

This week, Hewlett Packard Enterprise (HPE: NYSE) earnings will shed light on the state of AI infrastructure investment. But the more telling signal may come from cybersecurity. After an upbeat report from Palo Alto Networks (PANW: NASDAQ) and a sour one from Okta (OKTA: NASDAQ), all eyes turn to CrowdStrike (CRWD: NASDAQ). Their results will either confirm a bifurcated market—or expose something more systemic in cybersecurity procurement.

Last week, Dell (DELL: NYSE), Elastic (ESTC: NYSE), and Nvidia (NVDA: NASDAQ) each reported—though only one of them dictated the narrative. Blackwell is shipping. Inventories are lean. Supply is tightening. You don’t need a metaphor—just a calculator (links to our YouTube research reports are here, here, and here).

As always, I’m focused on three things:
1) Technology-driven change;
2) the latest in innovation and startup trends, and;
3) stock fraud.

Companies Discussed

In This Note:

• Companies Discussed

• Tempus Spins the Revenue Wheel

• The $2,093 iPhone: Six Reasons Apple Isn’t Made in …

• Tweet O’ The Week

• Epistrophy In The News

• 📆 of Epistrophy Events

• Availability This Week

Tempus Spins the Revenue Wheel

A month ago, we noted that Tempus AI (TEM: NASDAQ) was propping up its revenue with financial maneuvers—not just product demand. Last week, Spruce Point Capital dropped a 100-page report accusing the company of round-tripped deals, AI puffery, and related-party transactions. The stock promptly fell 19% (full disclosure, we have a personal short position in the stock).

We’ve received a bunch of questions about Tempus’ “$200 million” headline deal with AstraZeneca (AZN: NASDAQ) and Pathos AI — to us it’s less than it seems. The deal may allow Tempus to book revenue from both parties—while paying one with money from the other.

The deal’s structure is convoluted by design. Pathos—founded by Lefkofsky and helmed by Tempus insiders—agreed to pay Tempus $200 million over three years for data access. But up to half can be paid in Pathos Series D preferred stock. The rest comes in installments at unspecified increments.

On day one, Pathos paid Tempus $50 million as Tempus paid Pathos $35 million. Simultaneously, AstraZeneca paid Tempus $35 million. Does Pathos recognize that $35 million as revenue? Does Tempus recognize $85 million in revenue for a two-way circular transaction? Tempus isn’t clear.

What is clear: this financial engineering is driven only by related-party interest. It’s bookkeeping theater — and not the opening act. AstraZeneca appears to be beating a quiet retreat from a 2021 deal with Tempus. It has restructured its deal repeatedly. In 2024, Tempus recognized $16.3 million by reversing a contract asset tied to canceled AstraZeneca warrants. In December 2024, AstraZeneca let a $100 million warrant expire. The $35 million pass-through payment for work Pathos is meant to perform may satisfy part of that once-again-restructured Tempus deal. AstraZeneca seems to have reduced its liability. Tempus can book revenues.

Pathos raised $427 million in two rounds with investors include Lefkofsky’s Lightbank, firms tied to Tempus directors and Tempus itself. Some of that capital now circulates back to Tempus as revenue. 

Tempus AI has never made a dime (with an accumulated deficit of $2.2 billion). But its founder, Eric Lefkofsky — not one year after the IPO — has already cashed out more than $360 million in stock – his most recent sale just after the announcement and resulting stock surge from this circular “$200 million” deal.

The $2,093 iPhone: Six Reasons Apple Isn’t Made in America

The statement may be only eleven words long, but it implies hundreds of billions in capital costs, labor shifts and logistical disruptions.

Some might be expecting news about domestic iPhone manufacturing coming out of Apple’s (AAPL:NASDAQ) upcoming WWDC 2025. But we doubt it.

This report unpacks what something short of Trump’s “or else” would actually cost—and why making an iPhone in the U.S. is not just improbable but structurally infeasible in the short term.

Trump’s U.S.-Made iPhone Plan Could Raise Costs by 130%. Our analysis shows that manufacturing the iPhone in the United States—rather than China—would increase wholesale costs from $364.42 to $837.34, a 129.8 % jump. The Chinese-made iPhone 16 Pro retails for $1,199. To preserve Apple’s 60 % gross margin, a U.S.-made model would have to sell for $2,093.

Here are six reasons:

1. Labor: A 5x Cost Differential. President Trump’s May 23 demand that Apple shift iPhone production to the United States confronts a hard constraint: manufacturing economics. A device that costs $109 to build in China would cost $660 if assembled in the United States. That isn’t political speculation—it’s a supply chain fact, shaped by six measurable cost drivers.

The first is labor. Assembly work remains human-intensive. In China, line workers at Foxconn earn roughly $3.25 per hour including meals and housing. In India, wages hover around $2.00 per hour. In the United States, the Bureau of Labor Statistics reports average hourly compensation in electronics manufacturing at $26.18. This puts monthly labor cost per worker over $4,000 in the U.S., versus $780 in China. That difference alone adds over $230 per unit.

Table 1: Labor Cost, by region

2. Component Supply Chains: Geography Wins. China’s advantage extends beyond wages. Its supplier ecosystems are geographically concentrated. Most iPhone components—displays, batteries, sensors—are manufactured within a few hundred kilometers of Foxconn’s Zhengzhou plant. India is progressing, but not yet comparable. Apple began Indian assembly in 2017 with Wistron and expanded via Foxconn and Tata Electronics. By early 2025, India produced 18% of global iPhone volume, per India’s Ministry of Electronics & IT. However, many key inputs—including camera modules and SoCs—are still imported from China or Vietnam. Suppliers remain spread across Tamil Nadu, Karnataka, Maharashtra, and Uttar Pradesh.

Even with Apple’s capital and leverage, it has taken nearly eight years to reach partial localization in India. That timeline in a lower-cost, government-supported country underscores the scale of difficulty in attempting a similar U.S. shift from scratch.

3. CapEx: Building Costs Are 5.3x Higher Capital expenditure for final assembly in the U.S. is significantly higher. Foxconn’s Zhengzhou campus cost roughly $1 billion across several lines—approximately $100 million per line. Chinese construction workers earn $3–5 per hour, and permitting is fast-tracked. In contrast, similar U.S. facilities cost more than $530 million per line, based on data from TSMC’s Arizona fab and Samsung’s Taylor, Texas project. Union labor ($60–90/hour) and regulatory delays increase both costs and timelines. Amortized over 10 million units annually, this adds $64 per U.S.-made device versus $12 in China.

Table 2: Facility CapEx Amortization, by region

4. Compliance Costs: Red Tape Is Expensive The U.S. regulatory environment imposes additional cost. Compliance with OSHA, EPA, ADA, and state-level environmental rules requires legal staffing, audits, and build modifications. In contrast, permitting in China and India often proceeds through industrial development zones under expedited terms. We estimate that compliance costs add $97 per iPhone in the U.S., compared to $7 in China and $10 in India.

5. Logistics: Long Roads, Higher Freight Freight and logistics are structurally cheaper in China. Foxconn’s Zhengzhou facility has bonded rail and air links. Finished goods reach ports or airports at minimal internal cost. In India, some of this structure exists in Chennai and Bengaluru, but inter-regional shipping is slower. In the U.S., interstate trucking costs $2.75–$4.00 per mile, and congestion adds delay. Shipping phones from a Texas plant to U.S. coastal markets or ports adds ~$18 per unit. With warehousing and insurance, the total freight burden is ~$31 per phone in the U.S. versus $7 in China.

6. Workforce Development: No One’s Trained China has trained millions of line workers for electronics precision tasks. India has begun building that base with support from Apple and Tata. The U.S. lacks this workforce. Hiring, training, and onboarding electronics assemblers from scratch adds significant fixed cost and reduces early-line productivity. We estimate this adds $61 per unit in the U.S., compared to $2 in China.

Table 4: Manufacturing Cost, by region

For comparison, here’s what an iPhone would need to retail for in each country to maintain Apple’s 60% gross margin, factoring in per-device R&D, SG&A, and income taxes:

Table 4: iPhone 16 Pro Retail Price Breakdown at 60% Gross Margin

Conclusion: This Isn’t a Presidential-Term Project President Trump’s proposal could increase iPhone wholesale costs by 152%, according to our analysis—from $332.51 to $837.34. That scale of increase would reverberate through every aspect of Apple’s pricing and profitability.

Apple could, in theory, absorb a $500 cost increase per unit—but that would cut hardware margins by more than 60%, pushing them well below the 36.6% iPhone margin reported last quarter and forcing retail price hikes of over 75%.

Some third-party estimates suggest a $3,500 U.S.-made iPhone. Those numbers likely reflect full domestic supply chain replication, costly tariff assumptions, and no margin compression. Even under more conservative assumptions, our estimate—over $2,000 retail to maintain a 60% margin—renders the idea financially unworkable.

More importantly, this can’t be done quickly. India’s climb to 18% of global iPhone output took eight years with full government backing. The U.S. lacks dense supply chains, low-cost labor, or streamlined permitting. Assembling iPhones in the U.S. isn’t just expensive—it’s structurally out of reach in the near term. Until fundamentals shift, any reshoring demand is more political theater than operational plan.

Tweet O’ The Week

Epistrophy In The News

On NewsNation last week, I joined Connell McShane to explain why the Trump administration’s new crypto-in-retirement rules are a terrible idea. When your retirement portfolio can hold Dogecoin, you’re no longer investing—you’re just playing slots. We also discussed Nvidia’s remarkable quarter and why it’s incredible demand was actually understated by the quarter.

📆 of Epistrophy Events

Availability This Week

I’m available all week so email or even text if you don’t hear back right away. I’d love to expand on the thoughts I’ve share, and I’ll be right on top of all the earnings reports outlined above.  Written reports are available to clients, with video summaries on YouTube and, of course our popular summaries of the summaries (yes, the second derivative) on Instagram and Tiktok.

I hope these notes are helpful to you. I’d love to discuss them further and, as always, comments, questions and ideas are appreciated. If you have a friend or even a frenemy whom you think might benefit from this note, have them reach out and I’ll put them on the list.