A daily market update from FS Insight — what you need to know ahead of opening bell

“Unlike what Jensen would have you believe, Moore’s Law is alive and well.” — Intel CEO Pat Gelsinger, punching back at Nvidia boss Jensen Huang

Overnight

Global central banks plan to increase dollar reserves, survey suggests (FT)

Elon Musk orders Nvidia to ship thousands of AI chips reserved for Tesla to X and xAI (CNBC)

Tesla will likely spend ~$4B on Nvidia chips this year (RT)

HP enterprise earnings top estimates and guidance on strong AI server sales (Barron’s)

Narendra Modi’s party loses parliamentary majority in shock Indian election result (FT)

U.S. job openings fall to lowest level since February 2021 (YF)

Intel and Apollo in $11 billion joint venture for chip manufacturing plant in Ireland (WSJ)

Blackstone in UK rental housing push with £580mn Vistry deal (FT)

HPEjumped ~17% after beating Q2 earnings estimates and forecasting better-than-expected Q3 revenue thanks to upbeat demand for its AI servers (RT) 

BlackRock and Citadel Securities-backed group plan new national stock exchange in Texas (WSJ)

Chinese move billions to Hong Kong banks seeking higher yields (BBG)

Viswas Raghavan begins as Citigroup’s new head of IB (RT)

Citi boosts U.S. parental leave to compete with Wall Street rivals (BBG)

Larry Summer sees higher U.S. long-term rates over time (BBG)

Big Tech firm are unplugging stock market from reality (WSJ)

Swiss financial regulator wants to be able to name and shame banks (RT)

NYCB names CEO Joseph Otting as executive chairman (RT)

Power-hungry data centers spur U.S. talks with Big Tech (RT)

Major TikTok accounts hit by a cyberattack (RT)

Instagram is testing ‘unskippable’ ads (TC)

Google’s privacy chief to depart (RT) 

UAE seeks ‘marriage’ with US over AI deals (FT)

South Africa’s MultiChoice supports $2.9bn offer from France’s Canal+ (FT)

U.S. prosecutors probe global hacking-for-hire operation (WSJ)

Developers sit on empty lots after historic apartment boom (WSJ)

The dollar is at its strongest since the 1980s; can it last? (WSJ)

Cannabis producer seeks Boston Beer merger (WSJ)

Private equity puts brakes on healthcare roll-ups given government scrutiny (WSJ)

Influencers are driving a new category of unionizing: pharmacists (WSJ)

Why China’s overcapacity problem is about to get even worse, in seven charts (WSJ)

Jeffrey Katzenberg’s WndrCo announces closing of over $460 million for newest funds (BW)

Peter Orszag wants to reimagine Lazard. Will his bankers let him? (FT)

Nvidia and Salesforce double down on Canadian AI startup Cohere in $450 million round (YF)

Stock pickers defy Wall Street norm to risk it all on a few bets (BBG)

MrBeast hires new COO: ex-SoftBank Vision Fund managing partner who backed companies including DoorDash and oversaw Shutterfly IPO (TI)

Community-led consortium in Kenya’s tea-growing region battles to prevent a global PE firm from selling the renowned Lipton tea estates to a Sri Lankan conglomerate (Semafor)

Gro Intelligence, a Nairobi startup founded and run by ex-Morgan Stanley commodities trader and providing AI-powered climate-risk and food-production insights, shuts down (Semafor)

Chart of the Day

MARKET LEVELS

MORNING INSIGHT

Good morning!

What is notable in each of the last 3 trading sessions is the fact that stocks rally into the close. There is an old trader’s adage “amateurs open the market, professionals close the market” and thus, to us, the pattern in each of the last 3 days is a sign that institutions are re-risking:

Consider the sharp moves higher at end of day in each of the last 3 days:
– 5/31 +2% +72 points
– 6/3 +1% +56 points
– 6/4 +1% +53 points

To us, this is a sign of institutional accumulation, as ICI money market data shows institutions were de-risking in May (when S&P 500 rose +5%). Additionally, we sense that European macro investors are less constructive on U.S. equities, thus stocks seem to drift higher post-European close (~11am ET).

Click HERE for more.

TECHNICAL

Tuesday’s JOLTS data proved to be another catalyst which has coincided with weakness in Treasury yields and the U.S. Dollar and, conversely, strength in the stock market. Tuesday’s print of 8.059 million jobs missed expectations of 8.350 mm jobs expected, marking the lowest jobs level since February 2021.

Two, 5, 10 and 30-year yields pulled back sharply to levels right near May lows, which is the technical support barrier that, when violated, should result in a much sharper broad-based rally.

While the initial bad news failed to immediately prove to be good news for the U.S. stock market, prices pushed higher by Tuesday’s close, marking four straight sessions of closing well up off the lows following early weakness.

Click HERE for more.

CRYPTO

In our latest video we discuss how recent economic data is pushing the Fed Funds futures market in the right direction, the influence of rate expectations on the crypto market as observed through the Coinbase premium, the intersection of Bitcoin and AI, and why the near-term setup for COIN and STX is enticing.

Click HERE for more.

FIRST NEWS

Look for a Hallucinatory Search Experience, Coming Soon? In 2007 Google launched what the company termed ‘universal search’, incorporating images, books, and locations into their main search product: presenting de-contextualized answers based on data compiled by external providers who scarcely realized they were doing free R&D for Google. The move deprived content creators of the traffic they depended on, giving mass-produced, unverified information to users who never asked for it.

Google’s approach was less of a bet on accuracy of information than, say, TripAdvisor’s, FourSquare’s, or Yelp’s, and more of a bet on comprehensiveness. Thus Google could scrape a local competitor’s data, complete with inconsistencies (or incomplete) – and still deprive its competitors of a click if all the customers did was look at the search results page instead of clicking through.

Yet scale creates opportunities to address imperfections, and in this case it gave Google a faster, more automated feedback loop between mistaken data, passive customer feedback, and accuracy-focused updates. Today, over a billion active Google Maps users (many of whom have opted in to, or failed to opt out of, continuously sharing their location data with the company) collectively generate a dataset Google uses to offer users who search for a business’s operating hours its schedule and a histogram showing the average foot traffic at that location over time, while also showing how busy it is in real time.

Similarly, in AI, if Google takes its products seriously, they can improve them faster than anyone else because they get so much usage-based feedback. It’s well within Google’s financial and technological resources to take on the gargantuan Twitter dataset, check every single image to see whether it’s a screenshot of an AI-generated result, and either flag that result for improvement or mark it as a fake. In the meantime, Google is looking at how users interact with AI results, with regular results; how often they run the same search again using different phrasing, and so on.

For some categories, this approach has worked well. For instance, when Meta moved to short videos, it was also switching to a format in which users deliver the equivalent of either a thumbs-down or a partial thumbs-up a few times a minute by either continuing to watch their recommendations or switching to something else. When a few billion people are providing this sort of feedback, recommendation engines improve quickly.

The apocalyptic model in which AI tools eat all of the content and eliminate the economic incentive to make more of it is partly a failure of imagination, and partly an over-bet on what AI does well rather than what it does poorly. AI is good at telling you about the specific thing you’re missing once you’ve identified it, but quite bad at suggesting the concept you don’t realize you’re missing.

Thus, AI is a complement to explicitly defined uncertainty. As with search before it, once we have a tool, we are sure to find plenty of excuses to use it. ‘Near me’ searches were not something the original Google could support at all; it had a link graph, but not a concept of “a good place for a first date” or “best enchiladas within a 10 mile radius.” As Google evolved, so did the scope of questions you could ask it. AI expands that set further. If it’s a space as inexhaustible as our readiness to throw requests at DALL-E for increasingly idiosyncratic images or ask Claude ever-more-esoteric questions, then it will eventually take market share from other types of media. Still, these are tools, not entertainment.

In the long run, there’s a simple economic reason to expect more search queries to return generative answers: it’s easier to get perfectly optimized ad load producing every search result dynamically, made bespoke to that exact user, and to incorporate commercial messaging along the way. TheDiff