Last week’s Broadcom earnings were a reminder that scale itself can be a strategy. This week, Oracle, Synopsys and Rubrik take the stage, each with different stakes in the same story: how cloud software, design automation and security shape the next phase of computing and, yes, AI. If Broadcom made size the headline, these companies will show whether specialization can still be the edge.

DIG OUR COOL WEBSITE! We have a growing repository of past notes and a searchable research database at epistrophy.beehiiv.com.

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

• Chip Happens: Broadcom’s $10B XPU Play

  • Broadcom’s XPU Deal With OpenAI

  • Core Business Pillars

  • Broadcom as Barometer

• Ballmer’s Green Bet Up In Smoke

  • A climate tech fraud leads to an NBA scandal. 

• Tweet O’ The Week

• Epistrophy In The News

• 📆 of Epistrophy Events

• Availability This Week

Chip Happens: Broadcom’s $10B XPU Play

The semiconductor industry has never been short on spectacle, but Broadcom (AVGO: NASDAQ) has mastered the art of making scale itself the story. A company once best known for Ethernet chips now finds itself at the center of the largest capital shift in corporate history: the buildout of artificial intelligence infrastructure. When Broadcom reported Q3 results this week, the most important number wasn’t revenue or profit but the sheer proportion of business now tied to supplying the invisible plumbing of AI.

Broadcom reported $16.3 billion in quarterly revenue, up 25 percent from last year, and adjusted earnings per share of $1.74, beating Wall Street expectations. AI semiconductor sales came in at $5.3 billion, representing nearly one-third of total revenue and growing 65 percent year over year. Those figures extend a run of outperformance stretching back more than five years and underscore a profound structural shift. AI has gone from being a growth driver to becoming the organizing principle of Broadcom’s business.

Broadcom is not Nvidia (NVDA:NASDAQ). It doesn’t design GPUs or claim to reinvent computing every five years. Its role is quieter, but no less critical. Every GPU cluster relies on high-bandwidth connectivity, specialized accelerators and the intricate software–hardware integration that Broadcom has made its hallmark. If Nvidia builds the Ferrari, Broadcom paves the road and collects the tolls. That distinction explains why Wall Street now values Broadcom at nearly $640 billion, a valuation unimaginable only a few years ago.

This matters now because the AI boom is no longer hypothetical. Hyperscale data centers are consuming record levels of capital. Meta (META: NASDAQ) alone expects to spend nearly $40 billion on infrastructure this year. Amazon (AMZN: NASDAQ), Alphabet (GOOG: NASDAQ) and Microsoft (MSFT: NASDAQ) are not far behind. Every dollar they spend ripples through Broadcom’s ledgers. If the numbers show those dollars continue to swell, Broadcom confirms its position as indispensable. If not, the entire AI buildout story begins to wobble.

Broadcom’s XPU Deal With OpenAI

The most striking development this quarter wasn’t just the strong revenue gains but Broadcom’s announcement of a design partnership with an unnamed customer (shh… it’s OpenAI) to build custom XPUs—advanced accelerators that blend GPU-like parallelism with CPU-level flexibility and ASIC-style efficiency. Unlike Nvidia’s monolithic GPU model, XPUs are highly disaggregated: compute tiles, memory stacks, and interconnect fabric integrated via advanced packaging. OpenAI’s early prototypes reportedly rely on Broadcom’s co-packaging expertise to fuse HBM4 memory directly with compute dies using 2.5D interposers, driving bandwidth above 10 TB/s—critical for large-scale model training.

At the architectural level, these XPUs diverge from GPU-centric SIMT (single instruction, multiple thread) models. They feature a heterogeneous core mix tuned for transformer workloads: matrix-multiply engines tightly coupled with sparsity accelerators and inference-tuned vector cores. Broadcom’s ASIC team leads the RTL design, but the differentiator lies in packaging and interconnect. Using Broadcom’s 224G SerDes, the XPUs are expected to scale across racks with deterministic latency under 200 ns, effectively linking thousands of chips into a coherent fabric. Integration with OpenAI’s software stack—optimized partitioning across compute clusters—is what may give the platform an edge over generic GPUs.

What makes this deal transformative for Broadcom is positioning. For decades, it provided the “infrastructure underneath” AI. XPUs vault Broadcom into the compute plane itself. By co-designing not just the networking but the processing pipeline, Broadcom blurs the line between infrastructure and front-line compute vendor. If successful, XPUs would allow OpenAI to train frontier models with greater efficiency and at lower cost than GPU-bound hyperscalers, while simultaneously redefining Broadcom’s scope from enabler to primary architect of AI silicon ecosystems.

Core Business Pillars

Technically, Broadcom’s AI business still rests on three familiar pillars. The first is custom silicon for hyperscalers, where its ASIC division co-designs silicon for Google, Meta and Microsoft. The second is networking, where Broadcom switch ASICs like Tomahawk and Jericho define AI system topology. The third is VMware software, which stabilizes margins and amplifies infrastructure monetization. What the XPU deal introduces is a fourth pillar in the making—a credible, compute-centric revenue stream that could restructure Broadcom’s growth base for the next decade.

Broadcom as Barometer

Broadcom’s Q3 results put EBITDA margin at 67 percent, a level almost foreign to hardware businesses and more typical of entrenched software monopolies. It explains why the stock now trades at around 38 times forward earnings—valued as an indispensable platform, not a cyclical supplier. The durability test remains the same: whether AI infrastructure spending continues through power shortages, tariffs, and supply chain bottlenecks.

The trajectory is now clearer than ever. Broadcom has reached a point where its fortunes mirror those of AI itself. With the OpenAI XPU bet, it is no longer just the backbone of AI clusters—it is becoming the co-creator of AI compute itself. If GPUs were the foundation of the first stage of AI training, XPUs may prove the architecture of the second.

Ballmer’s Green Bet Up In Smoke

A climate tech fraud leads to an NBA scandal. 

Last week the Los Angeles Clippers grabbed headlines in an alleged financial fraud now under NBA and FBI investigation, focused on possible salary cap evasion and money laundering. At its heart is a venture-backed San Francisco “green tech/fintech” company that nearly pulled off a $2.3 billion IPO in 2021 before imploding through arrests, guilty pleas and the vaporization of nearly a billion dollars in investor funds.

Podcaster Pablo Torre went down the rabbit hole uncovering a byzantine story of LA Clippers owner and former Microsoft (MSFT:NASDAQ) CEO Steve Ballmer, NBA All Star Kawhi Leonard and alleged salary cap evasion (which Ballmer has denied). But underneath it all is a clean tech financial fraud called Aspiration Partners.

Aspiration Partners was once the toast of sustainable fintech. Founded in 2013 by former Clinton White House aides Andrei Cherny and Joseph Sanberg, they aimed to merge consumer banking and environmental stewardship. They built a brand on celebrity investors and marketing spectacle. By 2021, it had lined up a constellation of celebrity and billionaire backers— Orlando Bloom, Cindy Crawford, Leonardo DiCaprio, Robert Downey Jr., Drake, and Ballmer. They promoted the company’s climate-first branding and lending celebrity cachet to its ambitious fundraising rounds —and announced plans to go public through a SPAC merger at a $2.3 billion valuation. It forecast revenue of $254 million for 2022, more than double its $100.6 million 2021 top line.

What did the company sell? How about a Drake-branded green credit card? The “Aspiration Zero” credit cards rounded up purchases to fund the planting of trees— Aspiration claimed 15 million trees a year. The “Aspiration Plus” card promised to offset emissions for every gallon of gas purchased.

How about a zero-fee green energy fund? Aspiration’s “Redwood ESG Fund” was the vehicle that “can change climate change,” touting it as “100% fossil fuel free.” It marketed a “pay what is fair” fee model letting investors pick their annual management charge—except the fine print set a base fee of 0.5%, already significantly higher than typical passive ESG funds. The fund’s portfolio included supposedly green investments, but among its top holdings was Southwest Airlines (LUV:NYSE), a major consumer of jet fuel and a company rated “high ESG risk” for its environmental practices.

How about carbon credits? Cherny claimed that the company was planting “as many trees every day as there are in Central Park.” (Spoiler alert: they weren’t).  Aspiration would broker millions of tons of credits from forestry, methane capture and fossil fuel transition projects, ostensibly to facilitate corporate emission offsets. Clients included giants Meta (META:NASDAQ) and Microsoft. “Climate fees” were tacked on Clippers ticket prices to offset fan travel emissions.

The projects Aspiration sourced for credits—through its Catona Climate division—involved “nature-based carbon removal” (reforestation, agroforestry, sustainable agriculture), wind power, and similar schemes) sometimes with partner developers (e.g. scandal-ridden Adani Group’s wind project).

“Clean rich is the new filthy rich,” said Aspiration billboards on buildings in SoHo and Williamsburg – sporting what they claimed was NASA-developed “LumActiv paint” that “purifies the surrounding air of pollutants.”

The pitch sucked top venture capitalists from New York and Silicon Valley. Aspiration’s major venture capital and institutional backers included Alpha Edison (a leading Los Angeles VC), Capricorn Investment Group (a prominent Silicon Valley sustainability investor), Allen & Company (New York merchant bank/tech investor), Flourish Ventures (impact VC spun out from Omidyar Network), AGO Partners (early-stage venture fund), Deep Field Asset Management (hedge fund/VC), Social Impact Finance (impact investment fund), DNS Capital (Chicago-based family office of the Pritzker family), SuRo (SSSS:NASDAQ) (formerly GSV Capital, a publicly traded  VC), Radicle Impact (San Francisco impact/tech fund), and UBS O’Connor (the hedge fund arm of global bank UBS). Each of these investors participated in one or more of the company’s funding rounds from seed through Series C and venture debt.

But the SPAC filing apparently didn’t pass the sniff test. Aspiration’s claims were revealed to be more, well, aspirational than real. The company created a new metric it called “EBITDAM” — earnings before interest, taxes, depreciation, amortization and marketing—excluding the all important $149 million in 2021 marketing costs. For a business whose success hinged on relentless customer acquisition it was ridiculous.

Its five million customers were really just five million email addresses – with only about 592,000 accounts receiving funded deposits. 

Did they really plant the thirty-five million trees as so many filings and decks claimed? Torre’s podcast quoted an unnamed former Aspiration accountant saying Aspiration checked to see if any trees were ever planted only once.

The SPAC never launched. Subsequent struggles to secure bridge financing required multi-million dollar emergency packages, including an $18 million loan from secured bondholders.

In March 2025, Aspiration Partners, renamed “CTN Holdings,” filed for Chapter 11 in Delaware. The bankruptcy petition listed $170 million in liabilities, a $580 million net operating loss and urgent plans to sell assets within 45 days. Major creditors included Los Angeles Clippers and Kia Forum ($40 million, unsecured, for contracted carbon credits), Forum Entertainment ($11 million) – and Kawhi Leonard’s “KL2 Aspire LLC” ($28 million claim). Torre says Leonard never lifted a finger to promote Aspiration.

Board member Ibrahim AlHusseini of FullCycle, an early investor in Tesla (TSLA:NASDAQ), Bloom Energy (BE:NYSE), Uber (UBER:NASDAQ), and several other sustainability startupspled guilty to wire fraud after helping co-founder Joseph Sanberg falsify brokerage and bank statements to secure $145 million in loans, for which he received $12.3 million in fraudulent proceeds.

AlHusseini would plead guilty to wire fraud in February 2025; Sanberg was arrested days before Aspiration’s March 2025 bankruptcy filing. He pled guilty to wire fraud and securities fraud (sentencing pending as of September 2025) .The complaint alleges Sanberg inflated investor loan collateral by forging brokerage statements and bank documents—acquiring two major loans secured by shares of the company and deceiving investors about board member Ibrahim AlHusseini’s financial condition.

The NBA investigation is ongoing.

Tweet O’ The Week

Epistrophy In The News

On KCBS, I talked about the return of tech workers to San Francisco, the failures of BART and the unlikely resurgence of the ferry. On Schwab, I had a really fun hit with Sam Vadas where we had a facinating conversation on the antitrust implications facing Google — watch this one 👉 link. On NewsNation, I addressed how Washington is rethinking the role of big tech link. And on Yahoo Finance, I took a closer look at AI, chips and what Broadcom’s results tell us about both link.

📆 of Epistrophy Events

Availability This Week

I’ll be in San Francisco all week, a rare stretch without travel. It’s a chance to catch up on research, meet in person and follow the earnings parade.

Written reports are available to clients, with video summaries on YouTube, and of course our popular summaries of the summaries on Instagram and TikTok and YouTube Shorts.

I hope these notes are helpful to you. Suggestions? I’d love to discuss them further and, as always, comments, questions and ideas are appreciated.