undefined | Bet on an Nvidia rally – and do it on the cheap – with this strategy, according to the charts
Nvidia has delivered a spectacular run, climbing more than 1,200 % since October 2022 to become the largest component of the S&P 500 and the only stock with a market‑cap above $4 trillion—roughly the combined value of Broadcom, Meta Platforms and Tesla. The surge is driven by the company’s GPUs, especially the new Blackwell architecture and upcoming Rubin processors, which now power the majority of AI‑accelerated computing. With an 85 % share of the AI‑accelerator market and cloud giants pouring tens of billions into Nvidia hardware, the firm enjoys a near‑impenetrable moat in the artificial‑intelligence revolution, while still posting a forward P/E of about 22, which is surprisingly modest given its growth rate of roughly 60 % earnings per share.
What makes the current moment especially enticing is the pricing of Nvidia’s option market. The NVDA CallDex—a gauge of the cost of 30‑day out‑of‑the‑money call options—has collapsed to a reading of 55, the lowest since the end of 2021, after hovering above 200 in early 2024 and staying above 100 for much of the past two years. This sharp decline means speculative bullish positions are now far cheaper to enter. For example, with Nvidia trading around $183.60, a May 8, $200 call costs just $1.35 (about 0.7 % of the underlying price), breaking even at $201.35 and limiting loss to the premium paid.
For investors who believe Nvidia’s AI leadership will endure but are wary of buying the stock after such a massive rally, long call options offer a cost‑effective alternative. They provide defined risk, leverage upside potential, and—given today’s depressed pricing—present one of the most attractive volatility‑adjusted entry points since 2021. In short, while Nvidia’s dominance in AI hardware remains solid, the steep drop in call‑option prices creates a rare window for traders to gain exposure to the next leg of Nvidia’s story with far less capital and risk than purchasing the shares outright.
Read more: undefined
#nvidia #s&p500 #blackwell #nvdacalldex #ai-accelerator
Cursor (@cursor_ai)
Blackwell GPU에서 MoE 모델의 토큰 생성 방식을 재구성해 추론 속도를 1.84배 높이고 출력 정확도도 개선했다. 이 개선은 Composer 학습에도 직접 반영되어, 더 자주 향상된 모델 버전을 배포할 수 있게 한다.
OpenAIのo1モデル、NVIDIAのBlackwellで50倍の性能向上を実現
AI技術の急速な進化が続く中、OpenAIとNVIDIAの最新の取り組みが、驚異的な進化を見せている。OpenAIの最新“推論”モデル「o1」が、NVIDIAの次世代アーキテクチャ「Blackwell」を採用することで、推論能力を驚異的に向上させたことが明らかになった。 OpenAIのo1モデルとNVIDIAのBlackwellの相乗効果 OpenAIが最近発表した新しい大規模言語モデル（LLM）「o1」は、これまでの生成AIの枠を超え、「推論重視」の応答を生成することに焦点を当てている。この取り組みは、汎用人工知能の実現に向けた重要な一歩と言える。 一方、NVIDIAのCEOであるJensen […]https://xenospectrum.com/openais-o1-model-nvidias-blackwell-delivers-50x-performance-improvement/
NVIDIAのRTX 5090向け次世代GPU「GB202」はチップレットではなく巨大なモノリシック設計を維持か
NVIDIAが先日発表した「Blackwell」アーキテクチャとそれに基づくGB200 GPU等はNVIDIA初のMCM設計（チップレット設計）となったが、同じBlackwellアーキテクチャに基づくゲーミングGPUは少し趣が異なるようだ。リーカーのKopite7kimi氏によれば、GB202チップは巨大なモノリシックダイを採用すると言う。 GB202 […]
NVIDIA RTX 5090は448ビットバスと28GBのGDDR7メモリを搭載か、更なる「Super」や「Ti」登場の可能性も
NVIDIAの「Blackwell」RTX 5090グラフィックボードに搭載されるGB202 GPUとRTX 5080に搭載されるGB203がそれぞれテープアウトされたようだ。加えて、以前報じられたものとは異なり、RTX 5090のVRAM構成は少し控えめなものとなるという新たな情報ももたらされた。 RTX 5090はRTX 4090から50％の性能向上か Chiphell ForumのPanzerlied氏は、以前の情報を更新し、RTX […]
RTX 5080はあまり性能向上が見込めない？Blackwell GPUのGPC数に関する新たな情報が明らかに
NVIDIAのBlackwellゲーミングGPUのいくつかの詳細が、信頼できるリーカーのkopite7kimi氏からもたらされた。彼によれば、Blackwell GB20X系のGPC(Graphics Processing Cluster)数は、前モデルのAD10X系とほぼ同様な物になるようだ。 GPC数は変わらず、GB202はSM数が増えたが…… NVIDIAが今年発表するRTX […]
AISatoshi (@AiXsatoshi)
Opus 증류 데이터셋을 일본어화하기 위해 Blackwell GPU 4대를 돌리고 있다고 밝혔다. 대규모 GPU 인프라를 이용해 고속으로 데이터셋을 변환·생성하는 작업으로, 일본어용 학습 데이터 확장과 후속 튜닝에 유용한 흐름이다.
Are you tired of relying on cloud APIs for large language models?
The NVIDIA RTX Pro 6000 Blackwell is shipping, and it brings 96GB of GDDR7 ECC VRAM to a single workstation GPU. This means you can finally run 70B parameter models completely locally with room for massive context windows.
🔗 For read more visit the blog link: https://www.fitservers.com/blogs/nvidia-rtx-pro-6000-blackwell/
#Tech #Hardware #Nvidia #SelfHosted #MachineLearning #GenerativeAI #ServerAdmin #DataCenter #FitServers #Blackwell #OpenSourceAI
ServeTheHome
AllQuakes - EMSC
TheBadPlace
sayzard
Y Kobayashi
Fit Servers