MARA is a better buy now as an AI company than as a bitcoin company, and retiring its convertible notes was smart because it also removed the hedged shorts tied to them. The stock is up 35% since @qthomp called a death spiral imminent after MARA sold over $1B of bitcoin near $72,000. Miners have been unwinding bitcoin as a balance-sheet asset for over two years; public miners sold 60k BTC in 2022. x.com 

MARA's accelerating shift into AI infrastructure, plus upside from a bitcoin recovery, makes it attractive. Its $1.5 billion Long Ridge Energy & Power deal adds a 505 MW gas plant in Ohio, lifting power capacity 65% toward a 1GW+ AI campus. The Starwood venture targets 1 GW of IT capacity near term and over 2.5 GW long term, with tenant talks accelerating. Q1 revenue fell 18% on bitcoin prices; MARA sold $1.5 billion of bitcoin to cut debt and fund AI, still holding 50,000+ BTC. x.com 

MARA is finally selling its bitcoin reserves, which is the right move, but still needs to leave bitcoin mining, a failing business model. Per Lookonchain, MARA sold 23,093 BTC in the first half of 2026 and still holds 35,577. With the hashprice index at all-time lows, this is one of the most inefficient periods to mine in history. Using sale proceeds to pivot to AI/HPC could make MARA a really good opportunity in the coming weeks and months, if management has finally woken up. x.com 

Bitcoin appears near a local bottom: MARA's reported sale of 996 BTC is another contrarian bottom indicator. x.com 

MARA has outsized upside leverage to its infrastructure projects, an obvious pick for that trade. With Bitcoin at $60,000, a delay in either of its projects would create major funding issues, and that same sensitivity gives similar juice on the upside. x.com 

MARA's management is clearly clueless, having sold 25% of its bitcoin after a 25–30% drop while sitting years behind peers in the AI data-center transition. The company could be really undervalued if management got it together, but there are much better places for capital unless it cleans house or sells to an operator that can manage the assets better. A renewed bitcoin rally could warrant another look. x.com  x.com  x.com 

Bitdeer's mining business may support meaningful value even if Bitcoin stays near $80,000, giving its speculative AI case a second engine. Its cash-contribution chart assumes 15,720 BTC mined annually at $61,051 cash cost per BTC, implying $298M at $80,000 Bitcoin, versus a roughly $2.2B valuation. Making its own SEALMINER rigs gives it more control over technology, costs and expansion. This remains a speculative buy with substantial downside risk. x.com  x.com 

Cipher Mining's convertible notes cap intermediate-term upside at around $22. Raising $775M through an ATM while shares were in the $20s would have caused less dilution and avoided forward share-price risk; doing this deal at $11 instead was terribly timed. The stock is not worth touching here. x.com  x.com 

MARA lacks vision because CEO Fred Thiel keeps bouncing between strategies, from bitcoin treasury company to AI data-center play. Per @Tsuff101, he opposed the AI pivot and bought tens of thousands of bitcoin during the treasury hype, only for MARA to later sell 20K BTC around $60K, and now sounds more bullish on AI infrastructure than mining. He is late to the space, and his turning bullish is if anything a contrarian signal. x.com 

TeraWulf stands out among bitcoin miners pivoting to AI data centers because its massive 20-year Anthropic deal gives it real legitimacy, and some capacity is already energized, moving it beyond pure hype. Plenty of execution and buildout risk remains before that partnership produces, and the chart is losing its uptrend line, suggesting more short-term chop. Long term, it should do well if the buildout goes well and the AI revolution is real. x.com