CNBC The Exchange

2026-07-17 · Hosted by Kelly Evans · CNBC

Executive Summary

Kelly Evans led a wide-ranging episode on The Exchange, opening with GMO's Tom Hancock explaining why his firm has newly bought Nvidia — now trading under 20x forward earnings and "de-risked" — while avoiding memory stocks he believes are riding an unsustainable pricing spike. Wedbush's Egal Arunian laid out top picks (Alphabet, Uber, Reddit, Geometry) built around AI monetization and platform differentiation. Dallas Fed President Lorie Logan made an explicit call for the Fed to raise rates further, citing stalled progress on non-housing core services inflation. Venture capitalist Bradley Tusk discussed New York's first-in-the-nation data center moratorium and the bipartisan political backlash to AI's energy and social costs, while Wolf Research's Spencer Hannah revealed Claude ranks as the third most-used app among Wall Street interns, ahead of ChatGPT. The show closed with continued weakness in chips and memory (Sandisk, Micron, Seagate all down sharply) and Tim Seymour dissecting South Korea's leveraged ETF crackdown following extreme volatility in SK Hynix and Samsung.

Key Stories & Changes

1. Nvidia: The "De-Risked" AI Trade

  • Tom Hancock (GMO US Quality ETF): Nvidia has been "neglected" by hot money rotating into memory, now trading under 20x forward earnings and roughly flat year-to-date

  • Hancock's fund had avoided Nvidia previously in favor of semi-cap equipment names and Broadcom, but added it now given the multiple compression despite Nvidia remaining "the de facto standard in the industry"

  • Firm avoids memory stocks (DRAM, HBM) because pricing power reflects scarcity, not durable structural advantage — expects new capacity buildout in "a couple of years" to compress those record prices, similar to prior semiconductor cycles

  • Favors select industrials with AI/data-center power exposure: Schneider Electric and Trane (HVAC), both new holdings in Hancock's international quality ETF and a dedicated power infrastructure ETF (ticker: KWH)

2. Wedbush's Top AI-Adjacent Picks

  • Alphabet (top mega-cap pick): differentiated by its end-to-end platform and search monetization, which Egal Arunian says has "only gotten better" despite a year-old narrative about search's demise

  • Uber (large-cap pick, $91 price target): the recent Delivery Hero deal seen as unlocking international cross-sell, with cross-sell customers worth 3-4x more; also positioned as a partnership hub for autonomous vehicle networks rather than a standalone AV developer

  • Reddit (mid-cap pick): seen as monetizing its data asset via licensing deals with AI platforms as "real person" content becomes more valuable in an AI-saturated content landscape

  • Geometry (small-cap pick): a Siemens partnership (including a minority stake) seen as a durable moat and estimate-revision catalyst

3. Fed's Lorie Logan Calls Explicitly for Rate Hikes

  • Dallas Fed President Lorie Logan said "modestly higher interest rates" are necessary to bring inflation down, marking an escalation from her more hedged July 9th comments

  • Cited non-housing core services inflation stuck near 3% with "no progress in two years"

  • Market pricing: 5-year forward break-even inflation at 2.25%; futures market implies a 75% chance of a rate hike by December

  • Logan warned the Fed should not prioritize the employment mandate over price stability, warning that doing so historically necessitates larger future rate hikes

4. Data Center Backlash and New York's Moratorium

  • New York became the first state to pass a data center development moratorium; President Trump called the move "a terrible decision" on Truth Social

  • Bradley Tusk (Tusk Ventures): argued red-state/blue-state divergence won't hold because data centers must be near population centers regardless of politics; AI regulation is advancing in both red and blue states due to broad public anxiety

  • Evidence cited: in Georgia, 40% of the Public Utility Commission lost re-election after data center-driven price increases; similar consumer power-price impacts seen in Northern Virginia, Ohio, and Pennsylvania (PJM grid)

  • Tusk expects a workable path forward: data centers paying their own power/water costs (closed-loop systems, behind-the-meter generation), being better community partners, and possible bipartisan federal legislation if the House flips in the midterms

5. Claude Ranks Ahead of ChatGPT Among Wall Street Interns

  • Wolf Research's annual intern survey found Claude was the third most-used app among Gen Z interns at Wolf and partner firms — behind Instagram and Snapchat, but ahead of ChatGPT, which analyst Spencer Hannah said was a genuine surprise

  • Usage spans model building, written research, data crunching, and coding applications

  • TikTok ranked lower than expected in the same survey, even as it remains a fast-growing e-commerce channel (TikTok Shop) per Wolf's retail coverage

6. Anthropic IPO Timeline Moves Up

  • Bloomberg reported Anthropic's IPO could come as early as October, earlier than previously expected

7. Chip and Memory Rout Continues

  • Taiwan Semi, ASML, and Samsung all delivered good results that failed to impress investors

  • Sandisk, Micron, and Seagate all down sharply, most more than 30% from 52-week highs

  • SK Hynix down roughly 12% intraday, on pace for its worst month since 2008; had eight days of 5%+ swings in the first 12 trading days of July, including a 15% single-day drop — its worst day ever

  • South Korea's Financial Services Commission temporarily banned new leveraged single-stock ETF listings and raised minimum retail deposit requirements

  • Tim Seymour (Seymour Asset Management): SK Hynix and Samsung together typically represent 50-60% of the KOSPI, calling it a "hyper-concentrated" market; ~460,000 Korean accounts have already been liquidated via margin calls, mostly investors in their 20s-30s

  • Seymour favors rotating into Chinese tech (Alibaba, Baidu, Tencent) given cheaper AI accelerator chips and lower capex intensity relative to Korean/Taiwanese names

8. Housing Data Disappoints

  • Pending home sales fell 5.4% in June from May, down 0.3% year-over-year; 30-year mortgage rate at 6.64%

  • NAHB home builder sentiment fell to 34, below 50 for 15 consecutive months — the longest such stretch since 2012

  • Builders increasingly cutting prices (~6%) and raising incentives; existing home prices hit a record high in June

1. AI Investment Rotating From "Compute" to "Power" as the Binding Constraint

Multiple guests (Tom Hancock, Bradley Tusk) independently converged on electricity infrastructure — not chips — as the next critical bottleneck in AI buildout, driving both new investment themes (Schneider, Trane, power infrastructure ETFs) and political backlash (state moratoriums, utility commissioner losses).

2. Political Backlash to AI Becoming Bipartisan and Structural

Bradley Tusk's framing — that AI anxiety cuts across red and blue states alike, driven by fears over jobs, electricity costs, and children's wellbeing — suggests data center and AI regulation is becoming a durable policy front rather than a partisan one-off, with potential bipartisan federal legislation possible if political control shifts.

3. Differentiation, Not Scale, as the Winning AI Investment Thesis

Across Nvidia, hyperscalers, and Reddit, guests repeatedly emphasized that durable competitive advantage now comes from platform differentiation and lock-in (Microsoft's diverse product stack, Alphabet's vertical integration, Reddit's unique data) rather than raw infrastructure scale, which is increasingly viewed as commoditized.

4. South Korea's Concentration Risk as a Cautionary Tale for International Diversification

Tim Seymour's point that diversifying into international markets can simply trade one concentration risk (Mag 7) for another (Samsung/SK Hynix dominating the KOSPI) is a nuanced counterpoint to conventional "diversify internationally" advice. ---

Sentiment Analysis

Overall Market Sentiment: Selectively Constructive, Broadly Cautious on Chips

Guests were notably split: constructive to bullish on specific AI-adjacent equities (Nvidia, Alphabet, Uber) even as memory/chip names and Korean equities saw acute stress.

Risk Factors Highlighted

Memory pricing unsustainability: Scarcity-driven DRAM/HBM pricing seen as unlikely to hold once new capacity comes online.

South Korean margin-call contagion: ~460,000 liquidated accounts and record volatility in SK Hynix/Samsung threaten broader Asian tech sentiment.

Persistent core services inflation: Non-housing core services inflation stuck near 3% for two years, per Fed's Logan, risking further hikes.

Data center political backlash: State-level moratoriums and utility commissioner electoral losses signal a growing regulatory threat to AI infrastructure buildout.

LLM provider commoditization: Potential shakeout among LLM providers (Anthropic, OpenAI, etc.) could compress IPO valuations and pricing power.

Housing affordability deterioration: Home builder sentiment below 50 for 15 consecutive months, the longest stretch since 2012.

Anthropic IPO timing risk: Moved up to as early as October, raising execution and valuation uncertainty.

Hyperscaler compute constraints: Alphabet's reported $1B/month SpaceX compute rental highlights capacity bottlenecks even among the largest players.

This episode was covered in today's [The Market Signal — 2026-07-17](https://marketsignal.beehiiv.com/p/the-market-signal-2026-07-17), a cross-source synthesis of multiple podcast reports.

Keep Reading