Post

挖矿的小羊
挖矿的小羊
Show original
AMD财报炸了,股价却崩了——但真正的机会在Q4 AMD交出了一份“学霸级”财报:营收115.36亿美元,创历史新高,同比增长50%;数据中心业务67亿美元,同比暴增107%;调整后每股收益1.66美元,超市场预期。 然后盘后股价跌了9%。 你没看错,营收+50%、利润+246%、数据中心翻倍——盘后暴跌9%。 常规交易时段刚涨了7%,盘后就全吐出来还倒贴。 为什么? 因为市场要的不是“好”,是“惊艳”。 Q3营收指引130亿美元,已经高于分析师平均预期的125亿美元。但华尔街某些激进机构赌的是更高——130亿没满足他们。 资本开支飙到8.08亿美元,远高于市场预估的2.986亿美元。自由现金流环比下滑。 业绩超预期,股价照跌——这个剧本,我们币圈太熟了。 不就是“利好出尽即利空”吗? 今年以来AMD股价已经翻了一倍多,市场已经把“AI芯片挑战英伟达”这个故事提前定价了。现在你只是“好”,不够。 但我想说的是另一件事—— Q3的波动别太在意,Q4才是真正的翻盘点。 时间线梳理一下: 7月23日,苏姿丰在Advancing AI 2026大会宣布Helios全面量产。这是AMD首款机架级AI系统,直接对标英伟达的整柜方案。 8月4日,Q2财报确认:Helios Q3末开始出货,Q4放量。 首批客户:微软Azure、OpenAI、Meta、Anthropic、甲骨文。 重点来了—— Helios单机架集成72颗MI455X GPU,配备31TB HBM4内存。 Anthropic已承诺部署高达2GW的Helios系统。2GW什么概念?相当于2000兆瓦的AI算力部署,是AMD目前规模最大的AI基础设施合作之一。 OpenAI也规划2026年Q4将Helios上线,2027年加速部署。 这还不是全部—— AMD预测到2030年,AI加速器市场规模将达到1.4万亿美元。苏姿丰在财报电话会上表示,预计数据中心销售2027年将翻倍。 这不是画饼,这是有合同、有客户、有出货时间表的。 而本次财报——Helios没有任何收入贡献。 Q3末才开始出货的东西,Q2财报里怎么可能有? 所以,对加密世界来说,三个方向值得关注: 第一,AI算力租赁赛道。 Helios放量意味着更多AI算力进入市场。短期可能压低去中心化算力网络的定价,但长期看——更多算力供给会催生更多需求外溢,去中心化算力作为“补充算力”的价值反而可能提升。AMD公开的核心卖点不是“跑得更快”,而是“每美元推理token数最高多30%”——成本优势一旦确立,整个算力市场的定价逻辑都会重构。 第二,AI概念币的情绪催化。 如果Helios Q4出货超预期,AMD股价可能迎来重新定价,AI赛道加密资产可能联动。半导体板块和加密市场的资金联动已经不是秘密了——AMD盘后跌9%,Hyperliquid上AMD永续合约直接清算了206万美元。反过来也一样,利好来了就是共振。 第三,矿机替代逻辑。 更多算力流向AI,加密挖矿的硬件获取难度可能进一步上升。利好已有算力储备的矿企。 ⚠️ 但必须说清楚风险: Helios是Q3末才开始出货,本次财报没有任何Helios的收入贡献。 真正的业绩验证要在下次财报。 短期波动不改长期趋势,但别把Q3的“预期”当成Q2的“现实”来交易。 现在的AMD,就像2020年的英伟达——所有人都知道它牛,但没人确定它到底能牛到什么程度。估值已经打满,容错率几乎为零。 Q3的波动,是噪音。Q4的Helios放量,才是信号。 AI算力这条赛道,不是线性的。它是阶跃函数——每一次技术迭代,都是一次价格重估。 Helios就是AMD的阶跃点。 聪明钱在Q3布局,韭菜在Q4追高。 你自己选。 $BTC $AMD $XSPCX #AMD财报超预期,增长已被透支?
挖矿的小羊
挖矿的小羊
AMD released its Q2 2026 earnings report—— Revenue of $11.54 billion, up 50% year-over-year, a record high. Net profit of $2.3 billion, up 163% year-over-year. Data center revenue of $6.7 billion, up 107% year-over-year, accounting for 58% of total revenue. Earnings per share of $1.66, up 246% year-over-year. Revenue exceeded expectations, profit exceeded expectations, gross margin at 56%. All good news, right? Yet after hours, the stock price dropped 8.8%, once falling over 9%. A "high score" report was thrown into the trash by the market. There are only two reasons: spending too aggressively, expectations too high. First, capital expenditure was $808 million, while the market expected only $298.6 million. Nearly three times the difference. AMD’s burn rate on AI infrastructure has exceeded everyone’s imagination. Second, Q3 revenue guidance is $13 billion, although higher than analysts’ expectation of $12.5 billion, some buy-side institutions had previously bet on $14 billion. "Exceeding expectations" is no longer enough; the market wants "super exceeding expectations." Isn’t this exactly what we experience every day in the crypto world? Good news turns into bad news. The better the earnings, the harder the fall. Expectations are the only god in this market. So what does this mean for our crypto market? Two transmission chains directly related to your holdings. Chain One: GPUs are being snapped up by AI giants, miners are transforming into AI computing power providers AMD’s data center revenue doubled, driven mainly by demand for Instinct AI GPUs. GPU capacity is being fully booked by AI giants like Meta, OpenAI, and Oracle. It will only become harder and more costly for crypto miners to acquire new graphics cards. But an even more interesting thing is— Bitcoin miners are accelerating their transformation into AI computing power lessors. Public mining companies have already signed AI hosting contracts worth over $70 billion. Analysts expect that by the end of 2026, the proportion of AI revenue for listed mining companies may rise from the current 30% to 70%. Core Scientific just signed a 15-year, 529-megawatt AI infrastructure agreement with AMD, expected to bring in $14 billion in revenue. Riot Platforms’ data center business has already booked $33.2 million for the first time. Miners are no longer mining; they are becoming landlords for AI. What does this mean for decentralized GPU networks—Akash, Render? Traditional miners flooding into the AI leasing market with massive computing power are diluting the pricing power of decentralized computing power networks. The reuse of computing hardware is impacting the valuation logic of these projects. Chain Two: The "AMD moment" for AI concept tokens AMD dropped 9% after hours, triggering $2.06 million in liquidations on Hyperliquid’s AMD perpetual contracts, with longs accounting for 95%. Semiconductor sell-offs have repeatedly dragged Bitcoin down. Today BTC is consolidating around $64,000. If AMD continues to plunge after the US market officially opens tonight, AI-related crypto assets—FET, TAO, RNDR—are very likely to face correlated selling pressure. AI is the only sector in this cycle with earnings expectations. But every fluctuation in AI chip stocks will transmit to AI narrative tokens. Conversely, if the market digests the bad news and rebounds in a V-shape, that will be an opportunity to bottom-fish the AI sector. The key catalyst is still ahead: AMD’s first rack-level AI system Helios will start shipping to Meta, OpenAI, and Oracle this quarter. This is AMD’s first complete system integrating self-developed CPU, GPU, and network chips, directly competing with Nvidia’s full machine solutions. Helios’ Q3 shipment data will be the real decisive factor. Don’t die from the "guidance below expectations" just before dawn. $BTC $AMD $ETH #AMD财报超预期,增长已被透支?

Disclaimer: OKX Orbit content is provided for informational purposes only. Learn more

Replies

No comments yet. Be the first to reply!