Weekly Market Outlook 2026/06/08
Hello all,
Hope you all had a great weekend and got some much needed green light therapy after the last Friday.
Realtime Discord access with equity positions tracker(new website rolling out to select subscribers), intraday orderflow & option dealer summaries included with the paid newsletter membership . Click here to subscribe.
Friday we had a fruitful day where we were able to capture about half the downside.
After that, we tried to catch the bottom and after a couple stop outs were able to get a counter rotation off the lows.
FROM GREED TO FEAR?
One of the most beautiful thing about the markets is that it pretends to be irrational to a certain subset of participants all the time and perfectly rational to another subset.
Friday was the first “real” red day on Nasdaq and suddenly the entire tone of the market shifted.
The main cause for the large sell was a blockbuster NFP report, which would then cause the Fed to hike rates.
This was reflected in the long-end of the US treasuries as well. The long-end pricing reflects the sentiment that the Fed’s path and equilibrium rates are higher down the road.
Why is Nasdaq panicking?
Hyperscalers are taking on more debt in order to build out as many data centers as they can and if the FFR starts inching higher, their payback time period for the said assets increases due to the debt servicing costs going higher (Algo, please please stop just tell me long or short I’m already snoozing here)
Just this week, Alphabet issued fresh debt to the tune of $84 billion and Meta planned to do the same.
In a filing dated June 2, Alphabet said it now aimed to raise $18 billion through the sale of Class A and C shares and $16.75 billion from depositary shares. It had earlier planned to raise $30 billion through concurrent public offerings backed by investment banks, split evenly between the two.
The company’s plans to raise $10 billion through a private placement of shares to Berkshire Hathaway and another $40 billion at-the-market offering program in the third quarter remain unchanged.
The stock offerings are set to finalize on June 4 with the depositary shares closing a day later, the company said.
A couple weeks ago, we touched upon how Anthropic may not be profitable unless you slice it into a very small snapshot in time. It is not related to this but just a small reminder that private markets do not have the same scrutiny as public ones and you can get a rosy story more traction in private markets than public ones.
MIDTERM MADNESS
U.S. President Donald Trump told reporters on Friday that his team is looking into the idea of AI companies giving the American public a stake in their firms, adding that he planned to host a meeting with AI executives as soon as next week.
“There’s something very interesting about it, where it almost becomes a partnership with the American public,” he told reporters aboard Air Force One. “We’ll look into that.”
If this train of thought of Trump results into something or not is still debatable but American government nationalizing AI labs is not a good precedent in my opinion.
DO WE EVEN NEED FRONTIER MODELS
SWE-Bench of models show a really interesting trend.
You can get 90% of the top tier model intelligence for 30% of the cost.
I asked Deepseek which models are worth the price
AI output below
The standout value models are:
DeepSeek-V4-Flash-Max: 0.790, $0.14 / $0.28
Best cheap option. It is only 1.6 percentage points behind GPT-5.2 and 1.6 points behind Gemini 3.1 Pro, while being radically cheaper.MiniMax M2.5: 0.802, $0.30 / $1.20
Probably the best overall value. It scores basically the same as GPT-5.2, Kimi K2.6, Qwen3.7 Max, MiniMax M3, and DeepSeek Pro-Max, but costs far less than most of them.MiniMax M3: 0.805, $0.60 / $2.40
Excellent value. Slightly better than MiniMax M2.5, still cheap.DeepSeek-V4-Pro-Max: 0.806, $1.74 / $3.48
Best if you want near-top open model performance and still reasonable output pricing.Qwen3.7 Max: 0.804, $1.25 / $3.75
Good, but not clearly better value than MiniMax M3 or DeepSeek-V4-Pro-Max.
Models that look poor on value
GPT-5.2: 0.800, $1.75 / $14
Good score, bad value in this table. MiniMax M2.5, MiniMax M3, DeepSeek-V4-Pro-Max, Qwen3.7 Max, and Gemini 3.1 Pro all beat or match it on score. Several are far cheaper.
Claude Sonnet 4.6: 0.796, $3 / $15
Not compelling on this metric. It is below the 0.80 cluster and expensive.
Gemini 3.1 Pro: 0.806, $2.50 / $15
Strong score, but weak value versus DeepSeek-V4-Pro-Max or MiniMax M3.
Claude Opus 4.8 / 4.7
Worth it only if you need the best performance below Mythos. Not value buys.”
AI output ends
Since everyone always thinks they need flagship models, corporations end up forking a huge amount of dollars for likely similar code output.
The biggest risk to top AI labs is people moving to cheaper models en-masse and causing a cascading drop in paid daily users for them.
