Some of my thoughts about this report showed as following:
1. Intel still makes 80% of CPUs for personal computers and 94% of them for servers. Its estimated revenue of 2020 maybe about $75 billion USD, and its net income $20.7B.
2. Its major rivalry, AMD, has the estimated revenue of $9.47B, up from $6.73B a year earlier(40.7% growth, YoY) and estimated net income of $1.5B, up from $756M (98% growth,YoY).
It means that every dollar growth in revenue can bring 2.4 dollars growth in net income, which reflects the feature of high-rent-value business. The economic scale decides the gross profit margin.
And this is the reason why Intel enjoys higher marginal rent value than AMD.
3. Besides Intel's dominant area, it had failed in RF and mobile device industry such as its doomed WiMax chipsets and mobile phone SoCs. Now it is facing two major problems: 1) its manufacturing skill has been left behind of TSMC for one or two generations; and 2) its lack of machine-learning products.
In the history, Intel had lost its competition against then Japanese companies in the RAM industry. Intel waived its white flag and turned into the CPU industry in 1980s. However, such successful turning point has never shown again since the launch of its wifi chipset. I do not have enough confidence in its new M&A strategy in the AI field.
4. On the other hand, the TSMC's 2 nm manufacturing advance is not unreachable for Intel. I think the nature physical barriers, like the quantum tunnelling effect, will delay the progress of IC manufacturing after 2nm. Although I've heard and read many issues in its manufacturing team. I still believe Intel's outstanding engineers can get things done after all. It only takes time and money. Nonetheless, this is another big problem. After TSMC having enjoyed monopolistic profits for quite a long time, even Intel could catch up with TSMC after spending about couple years and billions dollars, the remaining rent value will be limited for Intel. I am not sure whether the necessary costs of competition would be covered or not, especially considering that Intel has lost its big client, Apple. Of course, the intense competition would erode TSMC's rent value, too. It's not a good news for TSMC's investors who recently got on the board at such high stock prices.
同時也有10000部Youtube影片,追蹤數超過2,910的網紅コバにゃんチャンネル,也在其Youtube影片中提到,...
「quantum tunnelling」的推薦目錄:
quantum tunnelling 在 DIGITIMES 科技網 Facebook 的最讚貼文
利用量子穿隧效應原理製造電晶體,比起CMOS製程不僅體積縮小3倍,製造時間更可縮短5倍。
quantum tunnelling 在 余海峯 David . 物理喵 phycat Facebook 的精選貼文
【科普文分享】澳物理學家首次成功測量量子穿隧時間/還在學習 - Edward Ho
//無論你將球擲向牆身多少次,球一樣會反彈回來。不過更微觀的量子世界,則是另一番景像。即使粒子衝向障礙,它們仍有機會抵達到另一邊。量子物理學家稱此現象為量子穿隧效應 (Quantum tunnelling effect)。
粒子可以穿過障礙是一回事,但需要花費多少時間呢?澳洲格里菲斯大學 (Griffith University) 量子物理學家 Igor Litvinyuk 及 Robert Sang 就用了三年時間,嘗試測量量子穿隧效應用到的時間。
在眾多元素中,他們選擇以氫原子 (atomic hydrogen) 作研究對象。透過量度極短頻光束數據,他們成功測量電子從氫原子中流離或走脫,以及穿隧延遲 (quantum delay) 的時間。他們發現粒子可在瞬間穿透障礙——有部份沒有參與研究的人甚至推論,粒子穿隧效應是否超越了光速。
不過,事實並非如此。《福布斯》專欄作家、天體物理學家 Ethan Siegel 解釋,雖然粒子穿越速度甚高,但其實仍比光速慢。他指出,量子世界的粒子活動不如經典物理世界般可被預測。假設你將粒子放在平面上,過一段時間再想找出其位置的話,礙於其不可測特性,現時無法透過移動速度和時間計算出實際位置,最多只可以機率解釋量子粒子活動。實際粒子的活動在未量度前都是機率,需在量度一刻才可肯定。另外,「粒子」本身也不是一顆顆般存在,而是類似波動般存在。
粒子轉移速度,會被障礙厚度、大小或其他未知的物理特性限制,也就是說轉移速度會在光速之下。在今次研究中,研究人員就透過光子量度電子轉移速度。他們發現在短短 1.8 原秒 (attoseconds) ,電子就穿越了障礙。相比之下,電子圍繞原子核一圈則需百多原秒。而在粒子穿隧過程中,假設只量度前段光束頻率,那就會出現「快過光速」的錯覺。然而,只要將所有光子也計算在內的話,其實電子根本沒快過光速。
今次研究首次在排除其他物理因素下,測量到單一粒子量子穿隧所需時間。實驗也同時反映在穿隧過程不會出現延遲時間,發現有助日後研發量子限制電晶體 (quantum-limited transitor) 。//